NCSC-TG-005 Library No. S228,526 Version 1 FOREWORD This publication is issued by the National Computer Security Center (NCSC) as part of its program to promulgate technical computer security guidelines. The interpretations extend the evaluation classes of the Trusted Systems Evalua- tion Criteria (DOD 5200.28-STD) to trusted network systems and components. This document will be used for a period of at least one year after date of signature. During this period the NCSC will gain experience using the Trusted Network Interpreta- tion in several network evaluations. In addition, the NCSC will conduct a series of tutorials and workshops to educate the community on the details of the Trusted Network Interpretation and receive feedback. After this trial period, necessary changes to the document will be made and a revised version issued. Workshops and tutorials will be open to any member of the network security community interested in providing feed- back. Anyone wishing more information, or wishing to pro- vide comments on the usefulness or correctness of the Trusted Network Interpretation may contact: Chief, Techni- cal Guidelines Division, National Computer Security Center, Ft. George G. Meade, MD 20755-6000, ATTN: C11. The tele- phone number is (301) 859-4452. ______________________________________________ PATRICK R. GALLAGHER, JR. 31 July 1987 Director National Computer Security Center ACKNOWLEDGMENT ______________ Acknowledgment is extended to the members of the Work- ing Group who produced this Interpretation. Members were: Alfred Arsenault, National Computer Security Center (Chair); Dr. Roger Schell, Gemini Computers; Stephen Walker, Trusted Information Systems; Dr. Marshall Abrams, MITRE; Dr. Jonathan Millen, MITRE; Leonard LaPadula, MITRE; Robert Morris, NCSC; and Jack Moskowitz, NCSC. Also due ack- nowledgement for their many inputs to this interpretation are Steve Padilla and William Shockley, Gemini Computers. Introduction ____________ I.1. Scope _ _ _____ Part I of this document provides interpretations of the Department of Defense Trusted Computer System Evaluation Criteria (TCSEC) (DOD-5200.28-STD), for trusted computer/communications network systems. The specific secu- rity feature, the assurance requirements, and the rating structure of the TCSEC are extended to networks of computers ranging from isolated local area networks to wide-area internetwork systems. Part II of this document describes a number of addi- tional security services (e.g., communications integrity, denial of service, transmission security) that arise in con- junction with networks. Those services available in specific network offerings, while inappropriate for the rigorous evaluation applied to TCSEC related feature and assurance requirements, may receive qualitative ratings. The TCSEC related feature and assurance requirements, and the additional security services described herein are intended for the evaluation of trusted network systems designed to protect a range of sensitive information. As such, they require that physical, administrative, pro- cedural, and related protection measures adequate to the sensitivity of the information being handled are already in place. It is possible, and indeed common practice, to operate a network in a secure manner at a single system high sensitivity level without meeting any trust related feature or assurance requirements described herein. The full range of physical and administrative security measures appropriate to the highest sensitivity level of information on the net- work must be in place in order to operate a trusted network as described in this Interpretation. It is important to note that this Interpretation does not describe all the security requirements that may be imposed on a network. Depending upon the particular environment, there may be communications security (COMSEC), emanations security, physical security, and other measures required. An environmental evaluation process, such as that described in the ``Computer Security Requirements--Guidance for Applying the DoD TCSEC in Specific Environments'' (CSC- STD-003-85), can be used to determine the level of trust required by specific system environments. Similar analyses are applicable to networks evaluated under these Interpreta- tions. I.2. Purpose _ _ _______ As with the TCSEC itself, this Interpretation has been prepared for the following purposes: 1. To provide a standard to manufacturers as to what security features and assurance levels to build into their new and planned, commercial network products in order to provide widely available systems that satisfy trust requirements for sensitive applica- tions 2. To provide a metric by which to evaluate the degree of trust that can be placed in a given network sys- tem for processing sensitive information 3. To provide a basis for specifying security require- ments in acquisition specifications. With respect to the second purpose for development of the criteria, i.e., providing a security evaluation metric, evaluations can be delineated into two types: an evaluation can be performed on a network product from a perspective that excludes the application environment, or an evaluation can be done to assess whether appropriate security measures have been taken to permit the system to be used operation- ally in a specific environment. The former type of evalua- tion is done by the National Computer Security Center through the Commercial Product Evaluation Process. The latter type of evaluation, those done for the pur- pose of assessing a system's security attributes with respect to a specific operational mission, is known as a certification evaluation. It must be understood that the completion of a formal product evaluation does not consti- tute certification or accreditation for the system to be used in any specific application environment. On the con- trary, the evaluation report only provides a trusted network system's evaluation rating along with supporting data describing the product system's strengths and weaknesses from a computer security point of view. The system security certification and the formal approval/accreditation pro- cedure, done in accordance with the applicable policies of the issuing agencies, must still be followed before a net- work can be approved for use in processing or handling clas- sified information. Designated Approving Authorities (DAAs) remain ultimately responsible for specifying the security of systems they accredit. This Interpretation can be used directly and indirectly in the certification process. Along with applicable policy, it can be used directly as technical guidance for evaluation of the total system and for specifying system security and certification requirements for new acquisitions. Where a system being evaluated for certification employs a product that has undergone a Commercial Product Evaluation, reports from that process will be used as input to the certification evaluation. Technical data will be furnished to designers, evaluators, and the DAAs to support their needs for making decisions. The fundamental computer security requirements as defined in the TCSEC apply to this Interpretation. I.3. Background _ _ __________ The term ``sponsor'' is used throughout this document to represent the individual or entity responsible for presenting a component or network system for evaluation. The sponsor may be a manufacturer, vendor, architect, developer, program manager, or related entity. A network system is the entire collection of hardware, firmware, and software necessary to provide a desired func- tionality. A component is any part of a system that, taken by itself, provides all or a portion of the total functionality required of a system. A component is recursively defined to be an individual unit, not useful to further subdivide, or a collection of components up to and including the entire sys- tem. Because the term integrity has been used in various contexts to denote specific aspects of an overall issue, it is important for the reader to understand the context in which the term is used in this document. Within Part I, as in the TCSEC itself, the use of this term is limited to (1) the correct operation of NTCB hardware/firmware and (2) pro- tection against unauthorized modification of labels and data. Specifically, all NTCBs that support sensitivity labels (viz., Divisions A and B) must, as detailed in the Label Integrity section of the TCSEC, protect the labels that represent the sensitivity of information (contained in objects) and the corresponding authorizations of users (with subjects as surrogates). In addition, for network systems with a defined data integrity policy, the NTCB must control the accesses of users that modify information-. As part of the NTCB itself, such integrity policies will be supported by access control mechanisms based on the identity of indi- viduals (for discretionary integrity) and/or sensitivity levels (for mandatory integrity). In contrast, within Part II the term integrity relates to the mechanisms for informa- tion transfer between distinct components. This communica- tions integrity includes the issues for correctness of mes- sage transmission, authentication of source/destination, data/control/protocol communication field correctness and related areas. _________________________ - See, for example, K. J. Biba, Integrity Considera- _________ _________ tions for Secure Computer Systems, MTR-3153, The MITRE _____ ___ ______ ________ _______ Corporation, Bedford, MA, June 1975. In many network environments, encryption can play an essential role in protecting sensitive information. In Part I of this document, encryption is referenced as a basis for providing data and label integrity assurance. In Part II, encryption is described as a tool for protecting data from compromise or modification attacks. Encryption algorithms and their implementation are outside the scope of this docu- ment. This document was prepared from a DoD perspective and requires NSA approval relative to the use of encryption. In other contexts, alternate approval authority may exist. As with the TCSEC itself, this is a reference document and is not intended to be used as a tutorial. The reader is assumed to be familiar with the background literature on computer security and communications networks=. Part II assumes a familiarity with the terminology used within ISO Security Services documentation. _________________________ = See, for example, M. D. Abrams and H. J. Podell, Tutorial: Computer and Network Security, IEEE Computer ________ ________ ___ _______ ________ Society Press, 1987. * ISO 7498/Part 2 - Security Architecture, ISO / TC ___ ____ ____ _ ________ ____________ 97 / SC 21 / N1528 / WG 1 Ad hoc group on Security, Project 97.21.18, September 1986. I.3.1. Trusted Computer System Evaluation Criteria _ _ _ _______ ________ ______ __________ ________ The DoD TCSEC was published in December 1985 to provide a means of evaluating specific security features and assurance requirements available in ``trusted commercially available automatic data processing (ADP) systems,'' referred to in this document as Automated Information System (AIS). The rating scale of the TCSEC extends from a rating which represents a minimally useful level of trust to one for ``state of the art'' features and assurance measures. These technical criteria guide system builders and evalua- tors in determining the level of trust required for specific systems. When combined with environmental guidelines, minimum security protection requirements, and appropriate accreditation decisions for specific installations can be determined. The philosophy of protection embodied in the TCSEC requires that the access of subjects (i.e., human users or processes acting on their behalf) to objects (i.e., containers of sensitive information) be mediated in accor- dance with an explicit and well defined security policy. In order to ensure strict compatibility between TCSEC evaluated AIS and networks and their components, and to avoid the possible evolution of incompatible evaluation cri- teria, Part I of this document has been specifically prepared as an INTERPRETATION of the TCSEC for networks. It is based entirely on the principles of the TCSEC, uses all TCSEC basic definitions, and introduces new concepts only where essential to understanding the TCSEC in a network con- text. Unless otherwise stated, the TCSEC requirements apply as written. The approach of interpreting the TCSEC for net- works in general has already been successfully employed in a number of specific complex network and AIS applications. There are several security policy models that may be used with the reference monitor concept. The Bell-LaPadula model is commonly used but is not mandated. Similarly for integrity policy, models such as Biba have been proposed but are not mandated. For this network interpretation, no specific access control policy is required; however, it is necessary that either a secrecy policy, an integrity policy, or both be specified for enforcement by the reference moni- tor. In the context of network systems, there are a number of additional security services that do not normally arise in individual AIS, and are not appropriate to the detailed feature and assurance evaluation prescribed by the TCSEC. These security services, which may or may not be available in specific network offerings, include provisions for com- munications security, denial of service, transmission secu- rity, and supportive primitives, such as encryption mechan- isms and protocols. Part II of this document describes these services and provides a qualitative means of evaluat- ing their effectiveness when provided. Evaluation of Part II offered services is dependent upon the results of the system's Part I evaluation or component's Appendix A evaluation. A Part II evaluation rating of good in a component or system which has a low Part I trust rating is of limited value. The sponsor must iden- tify which security services are offered by a system or com- ponent for evaluation against Part II. The evaluators will normally give a none, minimum, fair or good rating for those services offered. In some cases, a rating of present is all that can be given or a quantitative measure of strength may be used as the basis for rating. A not applicable rating will be given for services not offered by the product. Part II services may be provided by mechanisms outside the NTCB. I.3.2. Two Network Views _ _ _ ___ _______ _____ DoD Directive 5200.28 (and similar policies elsewhere in government) establishes the concept of a DAA, an indivi- dual who is responsible for approving the use of an AIS for processing classified information. For stand-alone AIS, this approval process and the technical advisory role to the DAA provided by the TCSEC are well understood. The same approval process applies to networks of AIS and plays a key role in determining how and when networks can be evaluated using this Interpretation. Depending upon the operational and technical charac- teristics of the environment in which a network exists, either of two views for accreditation and evaluation pur- poses applies: as a collection of two or more intercon- nected separately accredited AIS or as a single unified sys- tem the security accreditation of which is the responsibil- ity of a single authority. The security feature and assurance requirements of a specified network, and hence its suitability for evaluation under this Interpretation, is greatly impacted by which view of the network is appropriate. I.3.2.1. Interconnected Accredited AIS View _ _ _ _ ______________ __________ ___ ____ The interconnected accredited AIS view is an opera- tional perspective that recognizes that parts of the network may be independently created, managed, and accredited. Where different accrediting jurisdictions are involved, the joint approval process is required, describing the handling practices and classification levels that will be exchanged between the components involved. Interconnected accredited AIS consist of multiple sys- tems (some of which may be trusted) that have been indepen- dently assigned operational sensitivity ranges (the highest and lowest sensitivity levels of information that may be simultaneously processed on that system). In this view, the individual AIS may be thought of as ``devices'' with which neighboring systems can send and receive information. Each AIS is accredited to handle sensitive information at a sin- gle level or over a range between a minimum and maximum level. The range of sensitive information that may be exchanged between two such AIS is a range, agreed upon by each system's approving authorities, which cannot exceed the maximum sensitivity levels in common between the two sys- tems. Because of the complex structure of a network consist- ing of interconnected accredited AIS, it may not be practi- cal to evaluate such a network using this Interpretation or to assign it a trusted system rating. In this case, the accreditor is forced to accept the risk of assessing the security of the network without the benefit of an evaluation against the principles of the TCSEC as interpreted in Part I of the document. Appendix C describes the rules for con- necting separately accredited AIS and the circumstances in which these rules apply. I.3.2.2. Single Trusted System View _ _ _ _ ______ _______ ______ ____ The policy enforcement by trusted components in a ``single trusted system'' exhibits a common level of trust throughout. A single trusted system is accredited as a sin- gle entity by a single accrediting authority. (In certain circumstances where a system will process information from multiple sensitive sources, more then one accrediting authority may be involved, but their responsibility will be for accrediting the whole system as a single entity for use processing the information for which they have authority.) Networks such as these can be evaluated against this Interpretation and given a rating compatible with trusted AIS evaluated by the TCSEC itself. A ``single trusted system'' network implements a refer- ence monitor to enforce the access of subjects to objects in accordance with an explicit and well defined network secu- rity policy. The network has a single trusted computing base, referred to as the Network Trusted Computing Base (NTCB), which is partitioned (see section I.4.2) among the network components in a manner that ensures the overall net- work security policy is enforced by the network as a whole. Every component that is trusted must enforce a component-level security policy that may contain elements of the overall network security policy. The sum of all component-level security policies must be shown to enforce the overall network security policy. There is no requirement that every component in the network have an NTCB partition nor that any such partition comprise a complete TCB (e.g., a network component could be dedicated to supporting the audit function and implement only that portion of the NTCB). Interaction among NTCB par- titions shall be via communications channels, operating at either single or multiple levels as appropriate. The net- work security architect must identify how the NTCB is parti- tioned and how all the trusted system requirements are satisfied. A given component that does not enforce the full imple- mentation of all polices (i.e., mandatory access control, discretionary access control, identification/ authentication and audit) must be evaluated as a component as specified in Appendix A. For example, a network architecture that does not operate above Level 3 of the ISO protocol model and typ- ically does not enforce discretionary access control must be evaluated as a component under Appendix A and not as a full system. I.3.2.2.1. Connection-Oriented Abstraction _ _ _ _ _ __________ ________ ___________ In many networking environments, the overall network security policy includes controlling the establishment of authorized connections across the network. The access con- trol mediation performed by the components of these networks enforces the establishment of connections between host com- puters on the network in accordance with some form of authorized connection list. While a connection-oriented policy may be suitable from an overall network perspective, specifying such a policy in terms of component level abstractions may be difficult but is required in order to evaluate the network. Individual trusted network components may employ a local mechanism to enforce mediation only between local sub- jects and objects, as described in the TCSEC. Some of these components may have no direct involvement with the enforce- ment of network connections. Others, however, will have an additional higher level network connection enforcement role. This higher level connection-oriented abstraction may be enforced solely within an individual component or may be distributed across many components (e.g., in the end-to-end encryption case, cryptographic front end devices enforce the network connection authorization decisions made by an access control/key management center.) With the connection-oriented abstraction, the role of the network as a whole in controlling information flow may be more easily understood, but there may be no simple way to extend this abstraction to the reference monitor require- ments of individual components in the network. The overall network security policy must be decomposed into policy ele- ments that are allocated to appropriate components and used as the basis for security policy models for these com- ponents. The reference monitor subject/object definitions as given in the TCSEC represent the fundamental security policy enforcement at the individual component level but may not directly describe the overall network security policy issues such as the network's connection policy. The connection- oriented abstraction may be essential to understanding the overall network security policy. The network architecture must demonstrate the linkage between the connection-oriented abstraction and its realization in the individual components of the network. I.3.2.2.2. Subjects and Objects _ _ _ _ _ ________ ___ _______ For purposes of this trusted network interpretation, the terms ``subject'' and ``object'' are defined as in the TCSEC. The subjects of a trusted network commonly fall into two classes: those that serve as direct surrogates for a user (where ``user'' is synonymous with ``human being''), and ``internal'' subjects that provide services for other subjects--typically representing software process rather than being made part of each user surrogate subject. There is a set of TCSEC requirements that are directed at users, rather than subjects. In the network context, services used to facilitate communications between users and AIS (e.g., protocol handlers) are usually provided by inter- nal subjects. Some components that provide only communica- tions facilitating services have only internal subjects. Examples of ``single trusted system'' networks or com- ponents could include- packet-switched communications sub- networks (as found in the Defense Data Network (DDN), end- to-end (or host-to-host) encryption systems (such as used in Blacker or ANSI X9.17 implementations), application level networks or closed user community systems (such as the Inter Service/Agency Automated Message Processing Exchange [I S/A AMPE] and SACDIN Programs), local area networks, digital PABX systems, private switched networks (such as circuit- switched telecommunications systems), future Integrated Ser- vices Digital Network (ISDN) implementations, and a Virtual Machine Monitor (VMM) on a single computer when analyzed as a network. I.4. Evaluation of Networks _ _ __________ __ ________ The TCSEC provides a means for evaluating the trustworthiness of a system and assigning an evaluation class based on its technical properties - independent of the particular use for or the sensitivity of information being processed on the system. In this Interpretation, a network as a whole with its various interconnected components is recognized as a special instance of a trusted system. The designer of a trusted system is unconstrained by the TCSEC on design and implementation choices as long as for the _________________________ - Examples are employed throughout this document to clarify the concepts presented. The naming of an exam- ple implies no judgement of the product or system named nor on its suitability for any particular purpose. system as a whole there is a clearly distinguished TCB with a definitive protection domain boundary. The features and assurance measures provided within the TCB perimeter will determine the evaluation class. The network must be viewed as PARTITIONED into a set of interconnected components, where each component may have an independent ``NTCB parti- tion.'' All interaction between such trusted components must be via ``communication channels or I/O devices'' as defined by the TCSEC. For Division A and B networks these will generally be ``multilevel devices.'' I.4.1. Network Security Architecture and Design _ _ _ _______ ________ ____________ ___ ______ Any network evaluated under this Interpretation must possess a coherent Network Security Architecture and Design. (Interconnection of components that do not adhere to such a Network Security Architecture is addressed in the Intercon- nection Rules, Appendix C.) The Network Security Architec- ture must address the security-relevant policies, objec- tives, and protocols. The Network Security Design specifies the interfaces and services that must be incorporated into the network so that it can be evaluated as a trusted entity. There may be multiple designs that conform to the same architecture but which are more or less incompatible and non-interoperable (except through the Interconnection Rules). Security related mechanisms that require coopera- tion among components are specified in the design in terms of their visible interfaces; mechanisms which have no visi- ble interfaces are not specified in this document but are left as implementation decisions. The Network Security Architecture and Design must be available from the network sponsor before evaluation of the network, or any component, can be undertaken. The Network Security Architecture and Design must be sufficiently com- plete, unambiguous, and free from obvious flaws to permit the construction or assembly of a trusted network based on the structure it specifies. When a component is being designed or presented for evaluation, or when a network assembled from components is assembled or presented for evaluation, there must be a priori evidence that the Network Security Architecture and Design are satisfied. That is, the components are assembl- able into a network that conforms in every way with the Net- work Security Architecture and Design to produce a physical realization which is trusted to the extent that its evalua- tion indicates. In order for a trusted network to be constructed from components that can be built independently, the Network Security Architecture and Design must completely and unambi- guously define the security functionality of components as well as the interfaces between or among components. The Network Security Architecture and Design must be evaluated to determine that a network constructed to its specifica- tions will in fact be trusted, that is, it will be evaluatable under these Interpretations. I.4.2. The Partitioned NTCB _ _ _ ___ ___________ ____ Like a stand-alone system, the network as a whole possesses a single TCB, referred to as the NTCB, consisting of the totality of security-relevant portions of the net- work. But, unlike a stand-alone system, the design and evaluation of the network rests on an understanding of how the security mechanisms are distributed and allocated to various components, in such a way that the security policy is supported reliably in spite of (1) the vulnerability of the communication paths and (2) the concurrent, asynchronous operation of the network components. Some distributed systems have reliable, protected com- munication paths and thus satisfy only the first charac- teristic of a network: the division into concurrently operating, communicating processing components. Although certain interpretations in this Interpretation will not apply to them, it may be beneficial to employ this Interpre- tation to evaluate them, and to take advantage of the interpretations relating to component properties and inter- faces. An NTCB that is distributed over a number of network components is referred to as partitioned, and that part of the NTCB residing in a given component is referred to as an NTCB partition. A network host may possess a TCB that has previously been evaluated as a stand-alone system. Such a TCB does not necessarily coincide with the NTCB partition in the host, in the sense of having the same security perime- ter. Whether it does or not depends on whether the security policy for which the host TCB was evaluated coincides with the network security policy, to the extent that it is allo- cated to that host. Even when a network host has a TCB that has been previ- ously evaluated at a given class, and the host's TCB coin- cides with the host's NTCB partition, there is still no a priori relationship between the evaluation class of the host and the evaluation class of the network. Some examples will be given below to illustrate this point. To evaluate a network at a given class, each require- ment in Part I for that class must be satisfied by the net- work as a whole. It is also necessary to understand how each requirement is allocated among the network's com- ponents. Some components, such as the hosts, may satisfy the entire security policy in isolation; others, such as packet switches and access control centers, may have more specialized functions that satisfy only a subset of the net- work security policy. In addition, distinct subsets of the network security policy may be allocated to different net- work components. Forcing every component to satisfy a specific Part I requirement is neither necessary nor sufficient to ensure that the network as a whole meets that requirement. To show that it is not sufficient, consider two trusted multilevel AIS that export and import labeled information to and from each other over a direct connection. Both satisfy the Label Integrity requirement that a sensitivity label be accurately and unambiguously associated with exported data. If they were to have different, incompatible label encodings for the same sensitive information, the network as a whole would fail to satisfy the label integrity requirement. As a result, these Interpretations require at the B1 level and above that there be uniform labeling of sensitive informa- tion throughout the network. To show that it is not necessary, consider the Manda- tory Access Control requirement that at least two sensi- tivity levels be supported. Suppose that the network con- sists of a number of untrusted hosts that are incapable of maintaining labels and are operating at different levels in a single-level mode. If they are interconnected through suitable multilevel interface units, the network as a whole can support the ``two or more levels'' requirement. The allocation of a requirement to a component does not simply mean that the component satisfies the requirement in isolation, but includes the possibility that it depends on other components to satisfy the requirement locally, or cooperates with other components to ensure that the require- ment is satisfied elsewhere in the network. Taken together, these examples illustrate the essential role of an overall network security architecture in design- ing and evaluating a trusted network. I.4.3. Component Evaluation _ _ _ _________ __________ Because network components are often supplied by dif- ferent vendors and are designed to support standardized or common functions in a variety of networks, significant advantages can accrue from a procedure for evaluating indi- vidual components. The purpose of component evaluation is to aid both the network designer and the evaluator by per- forming the evaluation process once and reusing the results whenever that component is incorporated into a network. There are four types of security policies that may be supported by a network component: 1. Mandatory Access Control 2. Discretionary Access Control 3. Supportive policies (e.g., Authentication, Audit) 4. Application policies (e.g., the policy supported by a DBMS that is distinct from that supported by the underlying system) Application level policies are user dependent and will not be considered further in these Interpretations. For a component to support a policy such as Mandatory Access Controls, it must support all the required features for that policy with all of the required assurances of the given class. I.5. Structure of the Document _ _ _________ __ ___ ________ The remainder of this document is divided into two parts, three appendices, a list of acronyms, a glossary, and a list of references. Part I presents TCSEC statements and detailed interpretations, which together constitute the requirements against which networks will be evaluated; and rationale for the network interpretation of the TCSEC. The TCSEC statement applies as modified by the Interpretation. Part II is a description of other Security Services not covered in the TCSEC interpretation which may be applicable to networks. Appendix A describes the evaluation of network components. Appendix B describes the rationale for network partitioning into individual components. Appendix C describes the interconnect rules for linking interconnected accredited AIS. Part I: Interpretations of the ____ _ _______________ __ ___ Trusted Computer System Evaluation Criteria _______ ________ ______ __________ ________ Highlighting (ALL CAPS) is used in Part I to indicate criteria not contained in a lower class or changes and additions to already defined criteria. Where there is no highlighting, requirements have been carried over from lower classes without addition or modification. 1.0 DIVISION D: MINIMAL PROTECTION _ _ ________ _ _______ __________ This division contains only one class. It is reserved for those systems that have been evaluated but that fail to meet the requirements for a higher evaluation class. 2.0 DIVISION C: DISCRETIONARY PROTECTION _ _ ________ _ _____________ __________ Classes in this division provide for discretionary (need- to-know) protection and, through the inclusion of audit capabilities, for accountability of subjects and the actions they initiate. 2.1 CLASS (C1): DISCRETIONARY SECURITY PROTECTION _ _ _____ __ _____________ ________ __________ THE NETWORK TRUSTED COMPUTING BASE (NTCB) OF A CLASS (C1) NETWORK SYSTEM NOMINALLY SATISFIES THE DISCRETIONARY SECURITY REQUIREMENTS BY PROVIDING SEPARATION OF USERS AND DATA. IT INCORPORATES SOME FORM OF CREDIBLE CONTROLS CAPABLE OF ENFORC- ING ACCESS LIMITATIONS ON AN INDIVIDUAL BASIS, I.E., OSTENSIBLY SUITABLE FOR ALLOWING USERS TO BE ABLE TO PROTECT PRIVATE OR PROJECT INFORMATION AND TO KEEP OTHER USERS FROM ACCIDENTALLY READING OR DESTROYING THEIR DATA. THE CLASS (C1) ENVIRONMENT IS EXPECTED TO BE ONE OF COOPERATING USERS PRO- CESSING DATA AT THE SAME LEVEL(S) OF SENSITIVITY. THE FOLLOWING ARE MINIMAL REQUIREMENTS FOR SYSTEMS ASSIGNED A CLASS (C1) RATING. 2.1.1 Security Policy + Statement from DoD 5200.28-STD Implied from the Introduction to the TCSEC. + Interpretation THE NETWORK SPONSOR SHALL DESCRIBE THE OVERALL NETWORK SECURITY POLICY ENFORCED BY THE NTCB. AT A MINIMUM, THIS POLICY SHALL INCLUDE THE DISCRETIONARY REQUIREMENTS APPLICA- BLE TO THIS CLASS. THE POLICY MAY REQUIRE DATA SECRECY, OR DATA INTEGRITY, OR BOTH. THE POLICY SHALL INCLUDE A DISCRE- TIONARY POLICY FOR PROTECTING THE INFORMATION BEING PRO- CESSED BASED ON THE AUTHORIZATIONS OF USERS OR GROUPS OF USERS. THIS ACCESS CONTROL POLICY STATEMENT SHALL DESCRIBE THE REQUIREMENTS ON THE NETWORK TO PREVENT OR DETECT "READ- ING OR DESTROYING" SENSITIVE INFORMATION BY UNAUTHORIZED USERS OR ERRORS. UNAUTHORIZED USERS INCLUDE BOTH THOSE THAT ARE NOT AUTHORIZED TO USE THE NETWORK AT ALL (E.G., A USER ATTEMPTING TO USE A PASSIVE OR ACTIVE WIRE TAP) OR A LEGITI- MATE USER OF THE NETWORK WHO IS NOT AUTHORIZED TO ACCESS A SPECIFIC PIECE OF INFORMATION BEING PROTECTED. NOTE THAT "USERS" DOES NOT INCLUDE "OPERATORS," "SYSTEM PROGRAMMERS," "TECHNICAL CONTROL OFFICERS," "SYSTEM SECURITY OFFICERS," AND OTHER SYSTEM SUPPORT PERSONNEL. THEY ARE DISTINCT FROM USERS AND ARE SUBJECT TO THE TRUSTED FACILITY MANUAL AND THE SYSTEM ARCHITECTURE REQUIREMENTS. SUCH INDI- VIDUALS MAY CHANGE THE SYSTEM PARAMETERS OF THE NETWORK SYS- TEM, FOR EXAMPLE, BY DEFINING MEMBERSHIP OF A GROUP. THESE INDIVIDUALS MAY ALSO HAVE THE SEPARATE ROLE OF USERS. SECRECY POLICY: THE NETWORK SPONSOR SHALL DEFINE THE FORM OF THE DISCRETIONARY SECRECY POLICY THAT IS ENFORCED IN THE NETWORK TO PREVENT UNAUTHORIZED USERS FROM READING THE SENSITIVE INFORMATION ENTRUSTED TO THE NETWORK. DATA INTEGRITY POLICY: THE NETWORK SPONSOR SHALL DEFINE THE DISCRETIONARY INTEGRITY POLICY TO PREVENT UNAUTHORIZED USERS FROM MODIFYING, VIZ., WRITING, SENSITIVE INFORMATION. THE DEFINITION OF DATA INTEGRITY PRESENTED BY THE NETWORK SPONSOR REFERS TO THE REQUIREMENT THAT THE INFORMATION HAS NOT BEEN SUBJECTED TO UNAUTHORIZED MODIFICATION IN THE NET- WORK. + Rationale THE WORD "SPONSOR" IS USED IN PLACE OF ALTERNATIVES (SUCH AS "VENDOR," "ARCHITECT," "MANUFACTURER," AND "DEVELOPER") BECAUSE THE ALTERNATIVES INDICATE PEOPLE WHO MAY NOT BE AVAILABLE, INVOLVED, OR RELEVANT AT THE TIME THAT A NETWORK SYSTEM IS PROPOSED FOR EVALUATION. A TRUSTED NETWORK IS ABLE TO CONTROL BOTH THE READING AND WRITING OF SHARED SENSITIVE INFORMATION. CONTROL OF WRITING IS USED TO PROTECT AGAINST DESTRUCTION OF INFORMA- TION. A NETWORK NORMALLY IS EXPECTED TO HAVE POLICY REQUIRE- MENTS TO PROTECT BOTH THE SECRECY AND INTEGRITY OF THE INFORMATION ENTRUSTED TO IT. IN A NETWORK THE INTEGRITY IS FREQUENTLY AS IMPORTANT OR MORE IMPORTANT THAN THE SECRECY REQUIREMENTS. THEREFORE THE SECRECY AND/OR INTEGRITY POLICY TO BE ENFORCED BY THE NETWORK MUST BE STATED FOR EACH NET- WORK REGARDLESS OF ITS EVALUATION CLASS. THE ASSURANCE THAT THE POLICY IS FAITHFULLY ENFORCED IS REFLECTED IN THE EVALUATION CLASS OF THE NETWORK. THIS CONTROL OVER MODIFICATION IS TYPICALLY USED TO PROTECT INFORMATION SO THAT IT MAY BE RELIED UPON AND TO CONTROL THE POTENTIAL HARM THAT WOULD RESULT IF THE INFORMA- TION WERE CORRUPTED. THE OVERALL NETWORK POLICY REQUIRE- MENTS FOR INTEGRITY INCLUDES THE PROTECTION FOR DATA BOTH WHILE BEING PROCESSED IN A COMPONENT AND WHILE BEING TRANSMITTED IN THE NETWORK. THE ACCESS CONTROL POLICY ENFORCED BY THE NTCB RELATES TO THE ACCESS OF SUBJECTS TO OBJECTS WITHIN EACH COMPONENT. COMMUNICATIONS INTEGRITY ADDRESSED WITHIN PART II RELATES TO INFORMATION WHILE BEING TRANSMITTED. 2.1.1.1 Discretionary Access Control + Statement from DoD 5200.28-STD THE TCB SHALL DEFINE AND CONTROL ACCESS BETWEEN NAMED USERS AND NAMED OBJECTS (E.G., FILES AND PROGRAMS) IN THE ADP SYS- TEM. THE ENFORCEMENT MECHANISM (E.G., SELF/GROUP/PUBLIC CONTROLS, ACCESS CONTROL LISTS) SHALL ALLOW USERS TO SPECIFY AND CONTROL SHARING OF THOSE OBJECTS BY NAMED INDIVIDUALS OR DEFINED GROUPS OF INDIVIDUALS, OR BOTH. + Interpretation THE DISCRETIONARY ACCESS CONTROL (DAC) MECHANISM(S) MAY BE DISTRIBUTED OVER THE PARTITIONED NTCB IN VARIOUS WAYS. SOME PART, ALL, OR NONE OF THE DAC MAY BE IMPLEMENTED IN A GIVEN COMPONENT OF THE NETWORK SYSTEM. IN PARTICULAR, COM- PONENTS THAT SUPPORT ONLY INTERNAL SUBJECTS (I.E., THAT HAVE NO SUBJECTS ACTING AS DIRECT SURROGATES FOR USERS), SUCH AS A PUBLIC NETWORK PACKET SWITCH, MIGHT NOT IMPLEMENT THE DAC MECHANISM(S) DIRECTLY (E.G., THEY ARE UNLIKELY TO CONTAIN ACCESS CONTROL LISTS). IDENTIFICATION OF USERS BY GROUPS MAY BE ACHIEVED IN VARIOUS WAYS IN THE NETWORKING ENVIRONMENT. FOR EXAMPLE, THE NETWORK IDENTIFIERS (E.G., INTERNET ADDRESSES) FOR VARI- OUS COMPONENTS (E.G., HOSTS, GATEWAYS) CAN BE USED AS IDEN- TIFIERS OF GROUPS OF INDIVIDUAL USERS (E.G., "ALL USERS AT HOST A," "ALL USERS OF NETWORK Q") WITHOUT EXPLICIT IDENTIF- ICATION OF INDIVIDUAL USERS, NOR EVEN AN EXPLICIT NUMBER OF USERS IMPLIED), IF THIS IS CONSISTENT WITH THE NETWORK SECU- RITY POLICY. FOR NETWORKS, INDIVIDUAL HOSTS WILL IMPOSE NEED-TO-KNOW CONTROLS OVER THEIR USERS - MUCH LIKE (IN FACT, PROBABLY THE SAME) CONTROLS USED WHEN THERE IS NO NETWORK CONNECTION. WHEN GROUP IDENTIFIERS ARE ACCEPTABLE FOR ACCESS CON- TROL, THE IDENTIFIER OF SOME OTHER HOST MAY BE EMPLOYED, TO ELIMINATE THE MAINTENANCE THAT WOULD BE REQUIRED IF INDIVI- DUAL IDENTIFICATION OF REMOTE USERS WAS EMPLOYED. THE DAC MECHANISM OF A NTCB PARTITION MAY BE IMPLE- MENTED AT THE INTERFACE OF THE REFERENCE MONITOR OR MAY BE DISTRIBUTED IN SUBJECTS THAT ARE PART OF THE NTCB IN THE SAME OR DIFFERENT COMPONENT. THE REFERENCE MONITOR MANAGES ALL THE PHYSICAL RESOURCES OF THE SYSTEM AND FROM THEM CREATES THE ABSTRACTION OF SUBJECTS AND OBJECTS THAT IT CON- TROLS. SOME OF THESE SUBJECTS AND OBJECTS MAY BE USED TO IMPLEMENT A PART OF THE NTCB. WHEN INTEGRITY IS INCLUDED AS PART OF THE NETWORK DIS- CRETIONARY SECURITY POLICY, THE ABOVE INTERPRETATIONS SHALL BE SPECIFICALLY APPLIED TO THE CONTROLS OVER MODIFICATION, VIZ, THE WRITE MODE OF ACCESS, WITHIN EACH COMPONENT BASED ON IDENTIFIED USERS OR GROUPS OF USERS. + Rationale IN THIS CLASS, THE SUPPORTING ELEMENTS OF THE OVERALL DAC MECHANISM ARE TREATED EXACTLY AS UNTRUSTED SUBJECTS ARE TREATED WITH RESPECT TO DAC IN AN ADP SYSTEM, WITH THE SAME RESULT AS NOTED IN THE INTERPRETATION. STRENGTHENING OF THE DAC MECHANISM IN THE NETWORK ENVIRONMENT IS PROVIDED IN CLASS (C2) (SEE THE DISCRETIONARY ACCESS CONTROL SECTION). A TYPICAL SITUATION FOR DAC IS THAT A SURROGATE PROCESS FOR A REMOTE USER WILL BE CREATED IN SOME HOST FOR ACCESS TO OBJECTS UNDER THE CONTROL OF THE NTCB PARTITION WITHIN THAT HOST. THE INTERPRETATION REQUIRES THAT A USER IDENTIFIER BE ASSIGNED AND MAINTAINED FOR EACH SUCH PROCESS BY THE NTCB, SO THAT ACCESS BY A SURROGATE PROCESS IS SUBJECT TO ESSEN- TIALLY THE SAME DISCRETIONARY CONTROLS AS ACCESS BY A PRO- CESS ACTING ON BEHALF OF A LOCAL USER WOULD BE. HOWEVER, WITHIN THIS INTERPRETATION A RANGE OF POSSIBLE INTERPRETA- TIONS OF THE ASSIGNED USER IDENTIFICATION IS PERMITTED. THE MOST OBVIOUS SITUATION WOULD EXIST IF A GLOBAL DATABASE OF NETWORK USERS WERE TO BE MADE PERMANENTLY AVAIL- ABLE ON DEMAND TO EVERY HOST, (I.E., A NAME SERVER EXISTED) SO THAT ALL USER IDENTIFICATIONS WERE GLOBALLY MEANINGFUL. IT IS ALSO ACCEPTABLE, HOWEVER, FOR SOME NTCB PARTI- TIONS TO MAINTAIN A DATABASE OF LOCALLY-REGISTERED USERS FOR ITS OWN USE. IN SUCH A CASE, ONE COULD CHOOSE TO INHIBIT THE CREATION OF SURROGATE PROCESSES FOR LOCALLY UNREGISTERED USERS, OR (IF PERMITTED BY THE LOCAL POLICY) ALTERNATIVELY, TO PERMIT THE CREATION OF SURROGATE PROCESSES WITH PRESELECTED USER AND GROUP IDENTIFIERS WHICH, IN EFFECT, IDENTIFY THE PROCESS AS EXECUTING ON BEHALF OF A MEMBER OF A GROUP OF USERS ON A PARTICULAR REMOTE HOST. THE INTENT OF THE WORDS CONCERNING AUDIT IN THE INTERPRETATION IS TO PRO- VIDE A MINIMALLY ACCEPTABLE DEGREE OF AUDITABILITY FOR CASES SUCH AS THE LAST DESCRIBED. WHAT IS REQUIRED IS THAT THERE BE A CAPABILITY, USING THE AUDIT FACILITIES PROVIDED BY THE NETWORK NTCB PARTITIONS INVOLVED, TO DETERMINE WHO WAS LOGGED IN AT THE ACTUAL HOST OF THE GROUP OF REMOTE USERS AT THE TIME THE SURROGATE PROCESSING OCCURED. ASSOCIATING THE PROPER USER ID WITH A SURROGATE PROCESS IS THE JOB OF IDENTIFICATION AND AUTHENTICATION. THIS MEANS THAT DAC IS APPLIED LOCALLY, WITH RESPECT TO THE USER ID OF THE SURROGATE PROCESS. THE TRANSMISSION OF THE DATA BACK ACROSS THE NETWORK TO THE USER'S HOST, AND THE CREATION OF A COPY OF THE DATA THERE, IS NOT THE BUSINESS OF DAC. COMPONENTS THAT SUPPORT ONLY INTERNAL SUBJECTS IMPACT THE IMPLEMENTATION OF THE DAC BY PROVIDING SERVICES BY WHICH INFORMATION (E.G., A USER-ID) IS MADE AVAILABLE TO A COM- PONENT THAT MAKES A DAC DECISION. AN EXAMPLE OF THE LATTER WOULD BE THE CASE THAT A USER AT HOST A ATTEMPTS TO ACCESS A FILE AT HOST B. THE DAC DECISION MIGHT BE (AND USUALLY WOULD BE) MADE AT HOST B ON THE BASIS OF A USER-ID TRANSMIT- TED FROM HOST A TO HOST B. UNIQUE USER IDENTIFICATION MAY BE ACHIEVED BY A VARIETY OF MECHANISMS, INCLUDING (A) A REQUIREMENT FOR UNIQUE IDEN- TIFICATION AND AUTHENTICATION ON THE HOST WHERE ACCESS TAKES PLACE; (B) RECOGNITION OF FULLY QUALIFIED NETWORK ADDRESSES AUTHENTICATED BY ANOTHER HOST AND FORWARDED TO THE HOST WHERE ACCESS TAKES PLACE; OR (C) ADMINISTRATIVE SUPPORT OF A NETWORK-WIDE UNIQUE PERSONNEL IDENTIFIER THAT COULD BE AUTHENTICATED AND FORWARDED BY ANOTHER HOST AS IN (B) ABOVE, OR COULD BE AUTHENTICATED AND FORWARDED BY A DEDICATED NET- WORK IDENTIFICATION AND AUTHENTICATION SERVER. THE PROTO- COLS WHICH IMPLEMENT (B) OR (C) ARE SUBJECT TO THE SYSTEM ARCHITECTURE REQUIREMENTS. NETWORK SUPPORT FOR DAC MIGHT BE HANDLED IN OTHER WAYS THAN THAT DESCRIBED AS "TYPICAL" ABOVE. IN PARTICULAR, SOME FORM OF CENTRALIZED ACCESS CONTROL IS OFTEN PROPOSED. AN ACCESS CONTROL CENTER MAY MAKE ALL DECISIONS FOR DAC, OR IT MAY SHARE THE BURDEN WITH THE HOSTS BY CONTROLLING HOST-TO- HOST CONNECTIONS, AND LEAVING THE HOSTS TO DECIDE ON ACCESS TO THEIR OBJECTS BY USERS AT A LIMITED SET OF REMOTE HOSTS. IN THIS CASE THE ACCESS CONTROL CENTER PROVIDES THE LINKAGE BETWEEN THE CONNECTION ORIENTED ABSTRACTION (AS DISCUSSED IN THE INTRODUCTION) AND THE OVERALL NETWORK SECURITY POLICY FOR DAC. IN ALL CASES THE ENFORCEMENT OF THE DECISION MUST BE PROVIDED BY THE HOST WHERE THE OBJECT RESIDES. 2.1.2 Accountability 2.1.2.1 Identification and Authentication + Statement from DoD 5200.28-STD THE TCB SHALL REQUIRE USERS TO IDENTIFY THEMSELVES TO IT BEFORE BEGINNING TO PERFORM ANY OTHER ACTIONS THAT THE TCB IS EXPECTED TO MEDIATE. FURTHERMORE, THE TCB SHALL USE A PROTECTED MECHANISM (E.G., PASSWORDS) TO AUTHENTICATE THE USER'S IDENTITY. THE TCB SHALL PROTECT AUTHENTICATION DATA SO THAT IT CANNOT BE ACCESSED BY ANY UNAUTHORIZED USER. + Interpretation THE REQUIREMENT FOR IDENTIFICATION AND AUTHENTICATION OF USERS IS THE SAME FOR A NETWORK SYSTEM AS FOR AN ADP SYS- TEM. THE IDENTIFICATION AND AUTHENTICATION MAY BE DONE BY THE COMPONENT TO WHICH THE USER IS DIRECTLY CONNECTED OR SOME OTHER COMPONENT, SUCH AS AN IDENTIFICATION AND AUTHEN- TICATION SERVER. AVAILABLE TECHNIQUES, SUCH AS THOSE DESCRIBED IN THE PASSWORD GUIDELINE=, ARE GENERALLY ALSO APPLICABLE IN THE NETWORK CONTEXT. HOWEVER, IN CASES WHERE THE NTCB IS EXPECTED TO MEDIATE ACTIONS OF A HOST (OR OTHER NETWORK COMPONENT) THAT IS ACTING ON BEHALF OF A USER OR GROUP OF USERS, THE NTCB MAY EMPLOY IDENTIFICATION AND AUTHENTICATION OF THE HOST (OR OTHER COMPONENT) IN LIEU OF IDENTIFICATION AND AUTHENTICATION OF AN INDIVIDUAL USER. AUTHENTICATION INFORMATION, INCLUDING THE IDENTITY OF A USER (ONCE AUTHENTICATED) MAY BE PASSED FROM ONE COMPONENT TO ANOTHER WITHOUT REAUTHENTICATION, SO LONG AS THE NTCB PROTECTS (E.G., BY ENCRYPTION) THE INFORMATION FROM UNAU- THORIZED DISCLOSURE AND MODIFICATION. THIS PROTECTION SHALL PROVIDE AT LEAST A SIMILAR LEVEL OF ASSURANCE (OR STRENGTH OF MECHANISM) AS PERTAINS TO THE PROTECTION OF THE AUTHENTI- CATION MECHANISM AND AUTHENTICATION DATA. + Rationale THE NEED FOR ACCOUNTABILITY IS NOT CHANGED IN THE CON- TEXT OF A NETWORK SYSTEM. THE FACT THAT THE NTCB IS PARTI- TIONED OVER A SET OF COMPONENTS NEITHER REDUCES THE NEED NOR IMPOSES NEW REQUIREMENTS. THAT IS, INDIVIDUAL ACCOUNTABIL- ITY IS STILL THE OBJECTIVE. HOWEVER, IN THE CONTEXT OF A NETWORK SYSTEM AT THE (C1) LEVEL (WHEREIN EXPLICIT INDIVI- DUAL USER ACCOUNTABILITY IS NOT REQUIRED), "INDIVIDUAL ACCOUNTABILITY" CAN BE SATISFIED BY IDENTIFICATION OF A HOST (OR OTHER COMPONENT). IN ADDITION, THERE IS NO NEED IN A DISTRIBUTED PROCESSING SYSTEM LIKE A NETWORK TO REAUTHENTI- CATE A USER AT EACH POINT IN THE NETWORK WHERE A PROJECTION OF A USER (VIA THE SUBJECT OPERATING ON BEHALF OF THE USER) INTO ANOTHER REMOTE SUBJECT TAKES PLACE. THE PASSING OF IDENTIFIERS AND/OR AUTHENTICATION INFOR- MATION FROM ONE COMPONENT TO ANOTHER IS USUALLY DONE IN SUP- PORT TO THE IMPLEMENTATION OF THE DISCRETIONARY ACCESS CON- TROL (DAC). THIS SUPPORT RELATES DIRECTLY TO THE DAC REGARDING ACCESS BY A USER TO A STORAGE OBJECT IN A DIF- FERENT NTCB PARTITION THAN THE ONE WHERE THE USER WAS AUTHENTICATED. EMPLOYING A FORWARDED IDENTIFICATION IMPLIES ADDITIONAL RELIANCE ON THE SOURCE AND COMPONENTS ALONG THE PATH. 2.1.3 Assurance 2.1.3.1 Operational Assurance 2.1.3.1.1 System Architecture + Statement from DoD 5200.28-STD THE TCB SHALL MAINTAIN A DOMAIN FOR ITS OWN EXECUTION THAT PROTECTS IT FROM EXTERNAL INTERFERENCE OR TAMPERING (E.G., BY MODIFICATION OF ITS CODE OR DATA STRUCTURES). RESOURCES CONTROLLED BY THE TCB MAY BE A DEFINED SUBSET OF THE SUB- JECTS AND OBJECTS IN THE ADP SYSTEM. + Interpretation THE SYSTEM ARCHITECTURE CRITERION MUST BE MET INDIVIDU- ALLY BY ALL NTCB PARTITIONS. IMPLEMENTATION OF THE REQUIRE- MENT THAT THE NTCB MAINTAIN A DOMAIN FOR ITS OWN EXECUTION IS ACHIEVED BY HAVING EACH NTCB PARTITION MAINTAIN A DOMAIN FOR ITS OWN EXECUTION. THE SUBSET OF NETWORK RESOURCES OVER WHICH THE NTCB HAS CONTROL ARE THE UNION OF THE SETS OF RESOURCES OVER WHICH THE NTCB PARTITIONS HAVE CONTROL. CODE AND DATA STRUCTURES BELONGING TO THE NTCB, TRANSFERRED AMONG NTCB SUBJECTS (I.E., SUBJECTS OUTSIDE THE REFERENCE MONITOR BUT INSIDE THE NTCB) BELONGING TO DIFFERENT NTCB PARTITIONS, MUST BE PRO- TECTED AGAINST EXTERNAL INTERFERENCE OR TAMPERING. FOR EXAMPLE, A CRYPTOGRAPHIC CHECKSUM OR PHYSICAL MEANS MAY BE EMPLOYED TO PROTECT USER AUTHENTICATION DATA EXCHANGED BETWEEN NTCB PARTITIONS. + Rationale THE REQUIREMENT FOR THE PROTECTION OF COMMUNICATIONS BETWEEN NTCB PARTITIONS IS SPECIFICALLY DIRECTED TO SUBJECTS THAT ARE PART OF THE NTCB PARTITIONS. ANY REQUIREMENTS FOR SUCH PROTECTION FOR THE SUBJECTS THAT ARE OUTSIDE THE NTCB PARTITIONS ARE ADDRESSED IN RESPONSE TO THE INTEGRITY REQUIREMENTS OF THE SECURITY POLICY. 2.1.3.1.2 System Integrity + Statement from DoD 5200.28-STD HARDWARE AND/OR SOFTWARE FEATURES SHALL BE PROVIDED THAT CAN BE USED TO PERIODICALLY VALIDATE THE CORRECT OPERATION OF THE ON-SITE HARDWARE AND FIRMWARE ELEMENTS OF THE TCB. + Interpretation IMPLEMENTATION OF THE REQUIREMENT IS PARTLY ACHIEVED BY HAVING HARDWARE AND/OR SOFTWARE FEATURES THAT CAN BE USED TO PERIODICALLY VALIDATE THE CORRECT OPERATION OF THE HARDWARE AND FIRMWARE ELEMENTS OF EACH COMPONENT'S NTCB PARTITION. FEATURES SHALL ALSO BE PROVIDED TO VALIDATE THE IDENTITY AND CORRECT OPERATION OF A COMPONENT PRIOR TO ITS INCORPORATION IN THE NETWORK SYSTEM AND THROUGHOUT SYSTEM OPERATION. FOR EXAMPLE, A PROTOCOL COULD BE DESIGNED THAT ENABLES THE COM- PONENTS OF THE PARTITIONED NTCB TO EXCHANGE MESSAGES PERIOD- ICALLY AND VALIDATE EACH OTHER'S CORRECT RESPONSE. THE PRO- TOCOL SHALL BE ABLE TO DETERMINE THE REMOTE ENTITY'S ABILITY TO RESPOND. NTCB PARTITIONS SHALL PROVIDE THE CAPABILITY TO REPORT TO NETWORK ADMINISTRATIVE PERSONNEL THE FAILURES DETECTED IN OTHER NTCB PARTITIONS. INTERCOMPONENT PROTOCOLS IMPLEMENTED WITHIN A NTCB SHALL BE DESIGNED IN SUCH A WAY AS TO PROVIDE CORRECT OPERA- TION IN THE CASE OF FAILURES OF NETWORK COMMUNICATIONS OR INDIVIDUAL COMPONENTS. THE ALLOCATION OF DISCRETIONARY ACCESS CONTROL POLICY IN A NETWORK MAY REQUIRE COMMUNICATION BETWEEN TRUSTED SUBJECTS THAT ARE PART OF THE NTCB PARTI- TIONS IN DIFFERENT COMPONENTS. THIS COMMUNICATION IS NOR- MALLY IMPLEMENTED WITH A PROTOCOL BETWEEN THE SUBJECTS AS PEER ENTITIES. INCORRECT ACCESS WITHIN A COMPONENT SHALL NOT RESULT FROM FAILURE OF AN NTCB PARTITION TO COMMUNICATE WITH OTHER COMPONENTS. + Rationale THE FIRST PARAGRAPH OF THE INTERPRETATION IS A STRAIGHTFORWARD EXTENSION OF THE REQUIREMENT INTO THE CON- TEXT OF A NETWORK SYSTEM AND PARTITIONED NTCB AS DEFINED FOR THESE NETWORK CRITERIA. NTCB PROTOCOLS SHOULD BE ROBUST ENOUGH SO THAT THEY PERMIT THE SYSTEM TO OPERATE CORRECTLY IN THE CASE OF LOCAL- IZED FAILURE. THE PURPOSE OF THIS PROTECTION IS TO PRESERVE THE INTEGRITY OF THE NTCB ITSELF. IT IS NOT UNUSUAL FOR ONE OR MORE COMPONENTS IN A NETWORK TO BE INOPERATIVE AT ANY TIME, SO IT IS IMPORTANT TO MINIMIZE THE EFFECTS OF SUCH FAILURES ON THE REST OF THE NETWORK. ADDITIONAL INTEGRITY AND DENIAL OF SERVICE ISSUES ARE ADDRESSED IN PART II. 2.1.3.2 Life-Cycle Assurance 2.1.3.2.1 Security Testing + Statement from DoD 5200.28-STD THE SECURITY MECHANISMS OF THE ADP SYSTEM SHALL BE TESTED AND FOUND TO WORK AS CLAIMED IN THE SYSTEM DOCUMENTATION. TESTING SHALL BE DONE TO ASSURE THAT THERE ARE NO OBVIOUS WAYS FOR AN UNAUTHORIZED USER TO BYPASS OR OTHERWISE DEFEAT THE SECURITY PROTECTION MECHANISMS OF THE TCB. (SEE THE SECURITY TESTING GUIDELINES.) + Interpretation TESTING OF A COMPONENT WILL REQUIRE A TESTBED THAT EXERCISES THE INTERFACES AND PROTOCOLS OF THE COMPONENT. THE TESTING OF A SECURITY MECHANISM OF THE NETWORK SYSTEM FOR MEETING THIS CRITERION SHALL BE AN INTEGRATED TESTING PROCEDURE INVOLVING ALL COMPONENTS CONTAINING AN NTCB PARTI- TION THAT IMPLEMENT THE GIVEN MECHANISM. THIS INTEGRATED TESTING IS ADDITIONAL TO ANY INDIVIDUAL COMPONENT TESTS INVOLVED IN THE EVALUATION OF THE NETWORK SYSTEM. THE SPON- SOR SHOULD IDENTIFY THE ALLOWABLE SET OF CONFIGURATIONS INCLUDING THE SIZES OF THE NETWORKS. ANALYSIS OR TESTING PROCEDURES AND TOOLS SHALL BE AVAILABLE TO TEST THE LIMITS OF THESE CONFIGURATIONS. A CHANGE IN CONFIGURATION WITHIN THE ALLOWABLE SET OF CONFIGURATIONS DOES NOT REQUIRE RETEST- ING. + Rationale TESTING IS THE PRIMARY METHOD AVAILABLE IN THIS EVALUA- TION DIVISION TO GAIN ANY ASSURANCE THAT THE SECURITY MECHANISMS PERFORM THEIR INTENDED FUNCTION. 2.1.4 Documentation 2.1.4.1 Security Features User's Guide + Statement from DoD 5200.28-STD A SINGLE SUMMARY, CHAPTER, OR MANUAL IN USER DOCUMENTATION SHALL DESCRIBE THE PROTECTION MECHANISMS PROVIDED BY THE TCB, INTERPRETATIONS ON THEIR USE, AND HOW THEY INTERACT WITH ONE ANOTHER. + Interpretation THIS USER DOCUMENTATION DESCRIBES USER VISIBLE PROTEC- TION MECHANISMS AT THE GLOBAL (NETWORK SYSTEM) LEVEL AND AT THE USER INTERFACE OF EACH COMPONENT, AND THE INTERACTION AMONG THESE. + Rationale THE INTERPRETATION IS AN EXTENSION OF THE REQUIREMENT INTO THE CONTEXT OF A NETWORK SYSTEM AS DEFINED FOR THESE NETWORK CRITERIA. DOCUMENTATION OF PROTECTION MECHANISMS PROVIDED BY INDIVIDUAL COMPONENTS IS REQUIRED BY THE CRI- TERIA FOR TRUSTED COMPUTER SYSTEMS THAT ARE APPLIED AS APPROPRIATE FOR THE INDIVIDUAL COMPONENTS. 2.1.4.2 Trusted Facility Manual + Statement from DoD 5200.28-STD A MANUAL ADDRESSED TO THE ADP SYSTEM ADMINISTRATOR SHALL PRESENT CAUTIONS ABOUT FUNCTIONS AND PRIVILEGES THAT SHOULD BE CONTROLLED WHEN RUNNING A SECURE FACILITY. + Interpretation THIS MANUAL SHALL CONTAIN SPECIFICATIONS AND PROCEDURES TO ASSIST THE SYSTEM ADMINISTRATOR(S) MAINTAIN COGNIZANCE OF THE NETWORK CONFIGURATION. THESE SPECIFICATIONS AND PRO- CEDURES SHALL ADDRESS THE FOLLOWING: 1.THE HARDWARE CONFIGURATION OF THE NETWORK ITSELF; 2.THE IMPLICATIONS OF ATTACHING NEW COMPONENTS TO THE NETWORK; 3.THE CASE WHERE CERTAIN COMPONENTS MAY PERIODICALLY LEAVE THE NETWORK (E.G., BY CRASHING, OR BY BEING DISCONNECTED) AND THEN REJOIN; 4.NETWORK CONFIGURATION ASPECTS THAT CAN IMPACT THE SECURITY OF THE NETWORK SYSTEM; (FOR EXAMPLE, THE MANUAL SHOULD DESCRIBE FOR THE NETWORK SYSTEM ADMINISTRATOR THE INTERCONNECTIONS AMONG COMPONENTS THAT ARE CONSISTENT WITH THE OVERALL NETWORK SYSTEM ARCHITECTURE.) 5.LOADING OR MODIFYING NTCB SOFTWARE OR FIRMWARE (E.G., DOWN-LINE LOADING). THE PHYSICAL AND ADMINISTRATIVE ENVIRONMENTAL CONTROLS SHALL BE SPECIFIED. ANY ASSUMPTIONS ABOUT SECURITY OF A GIVEN NETWORK SHOULD BE CLEARLY STATED (E.G., THE FACT THAT ALL COMMUNICATIONS LINKS MUST BE PHYSICALLY PROTECTED TO A CERTAIN LEVEL). + Rationale THERE MAY BE MULTIPLE SYSTEM ADMINISTRATORS WITH DIVERSE RESPONSIBILITIES. THE TECHNICAL SECURITY MEASURES DESCRIBED BY THESE CRITERIA MUST BE USED IN CONJUNCTION WITH OTHER FORMS OF SECURITY IN ORDER TO ACHIEVE SECURITY OF THE NETWORK. ADDITIONAL FORMS INCLUDE ADMINISTRATIVE SECURITY, PHYSICAL SECURITY, EMANATIONS SECURITY, ETC. EXTENSION OF THIS CRITERION TO COVER CONFIGURATION ASPECTS OF THE NETWORK IS NEEDED BECAUSE, FOR EXAMPLE, PROPER INTERCONNECTION OF COMPONENTS IS TYPICALLY ESSENTIAL TO ACHIEVE A CORRECT REALIZATION OF THE NETWORK ARCHITEC- TURE. CRYPTOGRAPHY IS ONE COMMON MECHANISM EMPLOYED TO PRO- TECT COMMUNICATION CIRCUITS. ENCRYPTION TRANSFORMS THE REPRESENTATION OF INFORMATION SO THAT IT IS UNINTELLIGIBLE TO UNAUTHORIZED SUBJECTS. REFLECTING THIS TRANSFORMATION, THE SENSITIVITY OF THE CIPHERTEXT IS GENERALLY LOWER THAN THE CLEARTEXT. IF ENCRYPTION METHODOLOGIES ARE EMPLOYED, THEY SHALL BE APPROVED BY THE NATIONAL SECURITY AGENCY (NSA). THE ENCRYPTION ALGORITHM AND ITS IMPLEMENTATION ARE OUTSIDE THE SCOPE OF THESE INTERPRETATIONS. THIS ALGORITHM AND IMPLEMENTATION MAY BE IMPLEMENTED IN A SEPARATE DEVICE OR MAY BE A FUNCTION OF A SUBJECT IN A COMPONENT NOT DEDI- CATED TO ENCRYPTION. WITHOUT PREJUDICE, EITHER IMPLEMENTA- TION PACKAGING IS REFERRED TO AS AN ENCRYPTION MECHANISM HEREIN. 2.1.4.3 Test Documentation + Statement from DoD 5200.28-STD THE SYSTEM DEVELOPER SHALL PROVIDE TO THE EVALUATORS A DOCU- MENT THAT DESCRIBES THE TEST PLAN, TEST PROCEDURES THAT SHOW HOW THE SECURITY MECHANISMS WERE TESTED, AND RESULTS OF THE SECURITY MECHANISMS' FUNCTIONAL TESTING. + Interpretation THE "SYSTEM DEVELOPER" IS INTERPRETED AS "THE NETWORK SYSTEM SPONSOR". THE DESCRIPTION OF THE TEST PLAN SHOULD ESTABLISH THE CONTEXT IN WHICH THE TESTING WAS OR SHOULD BE CONDUCTED. THE DESCRIPTION SHOULD IDENTIFY ANY ADDITIONAL TEST COMPONENTS THAT ARE NOT PART OF THE SYSTEM BEING EVALUATED. THIS INCLUDES A DESCRIPTION OF THE TEST-RELEVANT FUNCTIONS OF SUCH TEST COMPONENTS AND A DESCRIPTION OF THE INTERFACING OF THOSE TEST COMPONENTS TO THE SYSTEM BEING EVALUATED. THE DESCRIPTION OF THE TEST PLAN SHOULD ALSO DEMONSTRATE THAT THE TESTS ADEQUATELY COVER THE NETWORK SECURITY POLICY. THE TESTS SHOULD INCLUDE THE FEATURES DESCRIBED IN THE SYSTEM ARCHITECTURE AND THE SYSTEM INTEGRITY SECTIONS. THE TESTS SHOULD ALSO INCLUDE NETWORK CONFIGURATION AND SIZING. + Rationale THE ENTITY BEING EVALUATED MAY BE A NETWORKING SUBSYS- TEM (SEE APPENDIX A) TO WHICH OTHER COMPONENTS MUST BE ADDED TO MAKE A COMPLETE NETWORK SYSTEM. IN THAT CASE, THIS INTERPRETATION IS EXTENDED TO INCLUDE CONTEXTUAL DEFINITION BECAUSE, AT EVALUATION TIME, IT IS NOT POSSIBLE TO VALIDATE THE TEST PLANS WITHOUT THE DESCRIPTION OF THE CONTEXT FOR TESTING THE NETWORKING SUBSYSTEM. 2.1.4.4 Design Documentation + Statement from DoD 5200.28-STD DOCUMENTATION SHALL BE AVAILABLE THAT PROVIDES A DESCRIPTION OF THE MANUFACTURER'S PHILOSOPHY OF PROTECTION AND AN EXPLA- NATION OF HOW THIS PHILOSOPHY IS TRANSLATED INTO THE TCB. IF THE TCB IS COMPOSED OF DISTINCT MODULES, THE INTERFACES BETWEEN THESE MODULES SHALL BE DESCRIBED. + Interpretation EXPLANATION OF HOW THE SPONSOR'S PHILOSOPHY OF PROTEC- TION IS TRANSLATED INTO THE NTCB SHALL INCLUDE A DESCRIPTION OF HOW THE NTCB IS PARTITIONED. THE SECURITY POLICY ALSO SHALL BE STATED. THE DESCRIPTION OF THE INTERFACES BETWEEN THE NTCB MODULES SHALL INCLUDE THE INTERFACE(S) BETWEEN NTCB PARTITIONS AND MODULES WITHIN THE PARTITIONS IF THE MODULES EXIST. THE SPONSOR SHALL DESCRIBE THE SECURITY ARCHITECTURE AND DESIGN, INCLUDING THE ALLOCATION OF SECURITY REQUIRE- MENTS AMONG COMPONENTS. APPENDIX A ADDRESSES COMPONENT EVALUATION ISSUES. + Rationale THE INTERPRETATION IS A STRAIGHTFORWARD EXTENSION OF THE REQUIREMENT INTO THE CONTEXT OF A NETWORK SYSTEM AS DEFINED FOR THIS NETWORK INTERPRETATION. OTHER DOCUMENTA- TION, SUCH AS DESCRIPTION OF COMPONENTS AND DESCRIPTION OF OPERATING ENVIRONMENT(S) IN WHICH THE NETWORKING SUBSYSTEM OR NETWORK SYSTEM IS DESIGNED TO FUNCTION, IS REQUIRED ELSE- WHERE, E.G., IN THE TRUSTED FACILITY MANUAL. IN ORDER TO BE EVALUATED, A NETWORK MUST POSSESS A COHERENT NETWORK SECURITY ARCHITECTURE AND DESIGN. (INTER- CONNECTION OF COMPONENTS THAT DO NOT ADHERE TO SUCH A SINGLE COHERENT NETWORK SECURITY ARCHITECTURE IS ADDRESSED IN THE INTERCONNECTION OF ACCREDITED AIS, APPENDIX C.) THE NETWORK SECURITY ARCHITECTURE MUST ADDRESS THE SECURITY-RELEVANT POLICIES, OBJECTIVES, AND PROTOCOLS. THE NETWORK SECURITY DESIGN SPECIFIES THE INTERFACES AND SERVICES THAT MUST BE INCORPORATED INTO THE NETWORK SO THAT IT CAN BE EVALUATED AS A TRUSTED ENTITY. THERE MAY BE MULTIPLE DESIGNS THAT CON- FORM TO THE SAME ARCHITECTURE BUT ARE MORE OR LESS INCOMPA- TIBLE AND NON-INTEROPERABLE (EXCEPT THROUGH THE INTERCONNEC- TION RULES). SECURITY RELATED MECHANISMS REQUIRING COOPERA- TION AMONG COMPONENTS ARE SPECIFIED IN THE DESIGN IN TERMS OF THEIR VISIBLE INTERFACES; MECHANISMS HAVING NO VISIBLE INTERFACES ARE NOT SPECIFIED IN THIS DOCUMENT BUT ARE LEFT AS IMPLEMENTATION DECISIONS. THE NETWORK SECURITY ARCHITECTURE AND DESIGN MUST BE AVAILABLE FROM THE NETWORK SPONSOR BEFORE EVALUATION OF THE NETWORK, OR ANY COMPONENT, CAN BE UNDERTAKEN. THE NETWORK SECURITY ARCHITECTURE AND DESIGN MUST BE SUFFICIENTLY COM- PLETE, UNAMBIGUOUS, AND FREE FROM OBVIOUS FLAWS TO PERMIT THE CONSTRUCTION OR ASSEMBLY OF A TRUSTED NETWORK BASED ON THE STRUCTURE IT SPECIFIES. WHEN A COMPONENT IS BEING DESIGNED OR PRESENTED FOR EVALUATION, OR WHEN A NETWORK ASSEMBLED FROM COMPONENTS IS ASSEMBLED OR PRESENTED FOR EVALUATION, THERE MUST BE A PRIORI EVIDENCE THAT THE NETWORK SECURITY ARCHITECTURE AND DESIGN ARE SATISFIED. THAT IS, THE COMPONENTS CAN BE ASSEM- BLED INTO A NETWORK THAT CONFORMS IN EVERY WAY WITH THE NET- WORK SECURITY ARCHITECTURE AND DESIGN TO PRODUCE A PHYSICAL REALIZATION THAT IS TRUSTED TO THE EXTENT THAT ITS EVALUA- TION INDICATES. IN ORDER FOR A TRUSTED NETWORK TO BE CONSTRUCTED FROM COMPONENTS THAT CAN BE BUILT INDEPENDENTLY, THE NETWORK SECURITY ARCHITECTURE AND DESIGN MUST COMPLETELY AND UNAMBI- GUOUSLY DEFINE THE SECURITY FUNCTIONALITY OF COMPONENTS AS WELL AS THE INTERFACES BETWEEN OR AMONG COMPONENTS. THE NETWORK SECURITY ARCHITECTURE AND DESIGN MUST BE EVALUATED TO DETERMINE THAT A NETWORK CONSTRUCTED TO ITS SPECIFICATIONS WILL IN FACT BE TRUSTED, THAT IS, IT WILL BE EVALUATABLE UNDER THESE INTERPRETATIONS. 2.2 CLASS (C2): CONTROLLED ACCESS PROTECTION _ _ _____ __ __________ ______ __________ NETWORK SYSTEMS IN THIS CLASS ENFORCE A MORE FINELY GRAINED DISCRETIONARY ACCESS CONTROL THAN (C1) NETWORK SYSTEMS, MAKING USERS INDIVIDUALLY ACCOUNTABLE FOR THEIR ACTIONS THROUGH LOGIN PRO- CEDURES, AUDITING OF SECURITY-RELEVANT EVENTS, AND RESOURCE ISOLATION. THE FOLLOWING ARE MINIMAL REQUIREMENTS FOR SYSTEMS ASSIGNED A CLASS (C2) RATING. 2.2.1 Security Policy _ _ _ ________ ______ + Statement from DoD 5200.28-STD Implied from the Introduction to the TCSEC. + Interpretation The network sponsor shall describe the overall network security policy enforced by the NTCB. At a minimum, this policy shall include the discretionary requirements applica- ble to this class. The policy may require data secrecy, or data integrity, or both. The policy shall include a discre- tionary policy for protecting the information being pro- cessed based on the authorizations of INDIVIDUALS, users, or groups of users. This access control policy statement shall describe the requirements on the network to prevent or detect "reading or destroying" sensitive information by unauthorized users or errors. Unauthorized users include both those that are not authorized to use the network at all (e.g., a user attempting to use a passive or active wire tap) or a legitimate user of the network who is not author- ized to access a specific piece of information being pro- tected. Note that "users" does not include "operators," "system programmers," "technical control officers," "system security officers," and other system support personnel. They are distinct from users and are subject to the Trusted Facility Manual and the System Architecture requirements. Such indi- viduals may change the system parameters of the network system, for example, by defining membership of a group. These individuals may also have the separate role of users. SECRECY POLICY: The network sponsor shall define the form of the discretionary secrecy policy that is enforced in the network to prevent unauthorized users from reading the sensitive information entrusted to the network. DATA INTEGRITY POLICY: The network sponsor shall define the discretionary integrity policy to prevent unauthorized users from modifying, viz., writing, sensitive information. The definition of data integrity presented by the network sponsor refers to the requirement that the information has not been subjected to unauthorized modification in the net- work. + Rationale The word "sponsor" is used in place of alternatives (such as "vendor," "architect," "manufacturer," and "developer") because the alternatives indicate people who may not be available, involved, or relevant at the time that a network system is proposed for evaluation. A trusted network is able to control both the reading and writing of shared sensitive information. Control of writing is used to protect against destruction of informa- tion. A network normally is expected to have policy require- ments to protect both the secrecy and integrity of the information entrusted to it. In a network the integrity is frequently as important or more important than the secrecy requirements. Therefore the secrecy and/or integrity policy to be enforced by the network must be stated for each net- work regardless of its evaluation class. The assurance that the policy is faithfully enforced is reflected in the evaluation class of the network. This control over modification is typically used to protect information so that it may be relied upon and to control the potential harm that would result if the informa- tion were corrupted. The overall network policy require- ments for integrity includes the protection for data both while being processed in a component and while being transmitted in the network. The access control policy enforced by the NTCB relates to the access of subjects to objects within each component. Communications integrity addressed within Part II relates to information while being transmitted. 2.2.1.1 Discretionary Access Control + Statement from DoD 5200.28-STD The TCB shall define and control access between named users and named objects (e.g., files and programs) in the ADP sys- tem. The enforcement mechanism (e.g., self/group/public controls, access control lists) shall allow users to specify and control sharing of those objects by named individuals or defined groups OF INDIVIDUALS, or both, AND SHALL PROVIDE CONTROLS TO LIMIT PROPAGATION OF ACCESS RIGHTS. THE DISCRE- TIONARY ACCESS CONTROL MECHANISM SHALL, EITHER BY EXPLICIT USER ACTION OR BY DEFAULT, PROVIDE THAT OBJECTS ARE PRO- TECTED FROM UNAUTHORIZED ACCESS. THESE ACCESS CONTROLS SHALL BE CAPABLE OF INCLUDING OR EXCLUDING ACCESS TO THE GRANULARITY OF A SINGLE USER. ACCESS PERMISSION TO AN OBJECT BY USERS NOT ALREADY POSSESSING ACCESS PERMISSION SHALL ONLY BE ASSIGNED BY AUTHORIZED USERS. + Interpretation The discretionary access control (DAC) mechanism(s) may be distributed over the partitioned NTCB in various ways. Some part, all, or none of the DAC may be implemented in a given component of the network system. In particular, com- ponents that support only internal subjects (i.e., that have no subjects acting as direct surrogates for users), such as a public network packet switch, might not implement the DAC mechanism(s) directly (e.g., they are unlikely to contain access control lists). Identification of users by groups may be achieved in various ways in the networking environment. For example, the network identifiers (e.g., internet addresses) for vari- ous components (e.g., hosts, gateways) can be used as iden- tifiers of groups of individual users (e.g., "all users at Host A," "all users of network Q") SO LONG AS THE INDIVIDU- ALS INVOLVED IN THE GROUP ARE IMPLIED BY THE GROUP IDENTIF- IER. FOR EXAMPLE, HOST A MIGHT EMPLOY A PARTICULAR GROUP-ID, FOR WHICH IT MAINTAINS A LIST OF EXPLICIT USERS IN THAT GROUP, IN ITS NETWORK EXCHANGE WITH HOST B, WHICH ACCEPTS THE GROUP-ID UNDER THE CONDITIONS OF THIS INTERPRETATION. For networks, individual hosts will impose need-to-know controls over their users ON THE BASIS OF NAMED INDIVIDUALS - much like (in fact, probably the same) controls used when there is no network connection. When group identifiers are acceptable for access con- trol, the identifier of some other host may be employed, to eliminate the maintenance that would be required if indivi- dual identification of remote users was employed. IN CLASS C2 AND HIGHER, HOWEVER, IT MUST BE POSSIBLE FROM THAT AUDIT RECORD TO IDENTIFY (IMMEDIATELY OR AT SOME LATER TIME) EXACTLY THE INDIVIDUALS REPRESENTED BY A GROUP IDENTIFIER AT THE TIME OF THE USE OF THAT IDENTIFIER. THERE IS ALLOWED TO BE AN UNCERTAINTY BECAUSE OF ELAPSED TIME BETWEEN CHANGES IN THE GROUP MEMBERSHIP AND THE ENFORCEMENT IN THE ACCESS CON- TROL MECHANISMS. The DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. The reference monitor manages all the physical resources of the system and from them creates the abstraction of subjects and objects that it con- trols. Some of these subjects and objects may be used to implement a part of the NTCB. WHEN THE DAC MECHANISM IS DISTRIBUTED IN SUCH NTCB SUBJECTS (I.E., WHEN OUTSIDE THE REFERENCE MONITOR), THE ASSURANCE REQUIREMENTS (SEE THE ASSURANCE SECTION) FOR THE DESIGN AND IMPLEMENTATION OF THE DAC SHALL BE THOSE OF CLASS C2 FOR ALL NETWORKS OF CLASS C2 OR ABOVE. When integrity is included as part of the network dis- cretionary security policy, the above interpretations shall be specifically applied to the controls over modification, viz, the write mode of access, within each component based on identified users or groups of users. + Rationale In this class, THE SUPPORTING ELEMENTS OF THE OVERALL DAC MECHANISM ARE REQUIRED TO ISOLATE INFORMATION (OBJECTS) THAT SUPPORTS DAC SO THAT IT IS SUBJECT TO AUDITING REQUIRE- MENTS (SEE THE SYSTEM ARCHITECTURE SECTION). THE USE OF NETWORK IDENTIFIERS TO IDENTIFY GROUPS OF INDIVIDUAL USERS COULD BE IMPLEMENTED, FOR EXAMPLE, AS AN X.25 COMMUNITY OF INTEREST IN THE NETWORK PROTOCOL LAYER (LAYER 3). IN ALL OTHER RESPECTS, the supporting elements of the overall DAC mechanism are treated exactly as untrusted subjects are treated with respect to DAC in an ADP system, with the same result as noted in the interpretation. A typical situation for DAC is that a surrogate process for a remote user will be created in some host for access to objects under the control of the NTCB partition within that host. The interpretation requires that a user identifier be assigned and maintained for each such process by the NTCB, so that access by a surrogate process is subject to essen- tially the same discretionary controls as access by a pro- cess acting on behalf of a local user would be. However, within this interpretation a range of possible interpreta- tions of the assigned user identification is permitted. The most obvious situation would exist if a global database of network users were to be made permanently avail- able on demand to every host, (i.e., a name server existed) so that all user identifications were globally meaningful. It is also acceptable, however, for some NTCB parti- tions to maintain a database of locally-registered users for its own use. In such a case, one could choose to inhibit the creation of surrogate processes for locally unregistered users, or (if permitted by the local policy) alternatively, to permit the creation of surrogate processes with preselected user and group identifiers which, in effect, identify the process as executing on behalf of a member of a group of users on a particular remote host. The intent of the words concerning audit in the interpretation is to pro- vide a minimally acceptable degree of auditability for cases such as the last described. What is required is that there be a capability, using the audit facilities provided by the network NTCB partitions involved, to determine who was logged in at the actual host of the group of remote users at the time the surrogate processing occured. Associating the proper user id with a surrogate process is the job of identification and authentication. This means that DAC is applied locally, with respect to the user id of the surrogate process. The transmission of the data back across the network to the user's host, and the creation of a copy of the data there, is not the business of DAC. Components that support only internal subjects impact the implementation of the DAC by providing services by which information (e.g., a user-id) is made available to a com- ponent that makes a DAC decision. An example of the latter would be the case that a user at Host A attempts to access a file at Host B. The DAC decision might be (and usually would be) made at Host B on the basis of a user-id transmit- ted from Host A to Host B. Unique user identification may be achieved by a variety of mechanisms, including (a) a requirement for unique iden- tification and authentication on the host where access takes place; (b) recognition of fully qualified network addresses authenticated by another host and forwarded to the host where access takes place; or (c) administrative support of a network-wide unique personnel identifier that could be authenticated and forwarded by another host as in (b) above, or could be authenticated and forwarded by a dedicated net- work identification and authentication server. The proto- cols which implement (b) or (c) are subject to the System Architecture requirements. Network support for DAC might be handled in other ways than that described as "typical" above. In particular, some form of centralized access control is often proposed. An access control center may make all decisions for DAC, or it may share the burden with the hosts by controlling host-to- host connections, and leaving the hosts to decide on access to their objects by users at a limited set of remote hosts. In this case the access control center provides the linkage between the connection oriented abstraction (as discussed in the Introduction) and the overall network security policy for DAC. In all cases the enforcement of the decision must be provided by the host where the object resides. 2.2.1.2 Object Reuse + Statement from DoD 5200.28-STD ALL AUTHORIZATIONS TO THE INFORMATION CONTAINED WITHIN A STORAGE OBJECT SHALL BE REVOKED PRIOR TO INITIAL ASSIGNMENT, ALLOCATION OR REALLOCATION TO A SUBJECT FROM THE TCB'S POOL OF UNUSED STORAGE OBJECTS. NO INFORMATION, INCLUDING ENCRYPTED REPRESENTATIONS OF INFORMATION, PRODUCED BY A PRIOR SUBJECT'S ACTIONS IS TO BE AVAILABLE TO ANY SUBJECT THAT OBTAINS ACCESS TO AN OBJECT THAT HAS BEEN RELEASED BACK TO THE SYSTEM. + Interpretation THE NTCB SHALL ENSURE THAT ANY STORAGE OBJECTS THAT IT CONTROLS (E.G., MESSAGE BUFFERS UNDER THE CONTROL OF A NTCB PARTITION IN A COMPONENT) CONTAIN NO INFORMATION FOR WHICH A SUBJECT IN THAT COMPONENT IS NOT AUTHORIZED BEFORE GRANTING ACCESS. THIS REQUIREMENT MUST BE ENFORCED BY EACH OF THE NTCB PARTITIONS. + Rationale IN A NETWORK SYSTEM, STORAGE OBJECTS OF INTEREST ARE THINGS THAT THE NTCB DIRECTLY CONTROLS, SUCH AS MESSAGE BUFFERS IN COMPONENTS. EACH COMPONENT OF THE NETWORK SYSTEM MUST ENFORCE THE OBJECT REUSE REQUIREMENT WITH RESPECT TO THE STORAGE OBJECTS OF INTEREST AS DETERMINED BY THE NETWORK SECURITY POLICY. FOR EXAMPLE, THE DAC REQUIREMENT IN THIS DIVISION LEADS TO THE REQUIREMENT HERE THAT MESSAGE BUFFERS BE UNDER THE CONTROL OF THE NTCB PARTITION. A BUFFER ASSIGNED TO AN INTERNAL SUBJECT MAY BE REUSED AT THE DISCRE- TION OF THAT SUBJECT WHICH IS RESPONSIBLE FOR PRESERVING THE INTEGRITY OF MESSAGE STREAMS. SUCH CONTROLLED OBJECTS MAY BE IMPLEMENTED IN PHYSICAL RESOURCES, SUCH AS BUFFERS, DISK SECTORS, TAPE SPACE, AND MAIN MEMORY, IN COMPONENTS SUCH AS NETWORK SWITCHES. 2.2.2 Accountability _ _ _ ______________ 2.2.2.1 Identification and Authentication + Statement from DoD 5200.28-STD The TCB shall require users to identify themselves to it before beginning to perform any other actions that the TCB is expected to mediate. Furthermore, the TCB shall use a protected mechanism (e.g., passwords) to authenticate the user's identity. The TCB shall protect authentication data so that it cannot be accessed by any unauthorized user. THE TCB SHALL BE ABLE TO ENFORCE INDIVIDUAL ACCOUNTABILITY BY PROVIDING THE CAPABILITY TO UNIQUELY IDENTIFY EACH INDIVI- DUAL ADP SYSTEM USER. THE TCB SHALL ALSO PROVIDE THE CAPA- BILITY OF ASSOCIATING THIS IDENTITY WITH ALL AUDITABLE ACTIONS TAKEN BY THAT INDIVIDUAL. + Interpretation The requirement for identification and authentication of users is the same for a network system as for an ADP sys- tem. The identification and authentication may be done by the component to which the user is directly connected or some other component, such as an identification and authen- tication server. Available techniques, such as those described in the Password Guideline=, are generally also applicable in the network context. However, in cases where the NTCB is expected to mediate actions of a host (or other network component) that is acting on behalf of a user or group of users, the NTCB may employ identification and authentication of the host (or other component) in lieu of identification and authentication of an individual user, SO LONG AS THE COMPONENT IDENTIFIER IMPLIES A LIST OF SPECIFIC USERS UNIQUELY ASSOCIATED WITH THE IDENTIFIER AT THE TIME OF ITS USE FOR AUTHENTICATION. THIS REQUIREMENT DOES NOT APPLY TO INTERNAL SUBJECTS. Authentication information, including the identity of a user (once authenticated) may be passed from one component to another without reauthentication, so long as the NTCB protects (e.g., by encryption) the information from unau- thorized disclosure and modification. This protection shall provide at least a similar level of assurance (or strength of mechanism) as pertains to the protection of the authenti- cation mechanism and authentication data. + Rationale The need for accountability is not changed in the con- text of a network system. The fact that the NTCB is parti- tioned over a set of components neither reduces the need nor imposes new requirements. That is, individual accountabil- ity is still the objective. ALSO, in the context of a net- work system at the (C2) LEVEL OR HIGHER "INDIVIDUAL ACCOUN- TABILITY" CAN BE SATISFIED BY IDENTIFICATION OF A HOST (OR OTHER COMPONENT) SO LONG AS THE REQUIREMENT FOR TRACEABILITY TO INDIVIDUAL USERS OR A SET OF SPECIFIC INDIVIDUAL USERS WITH ACTIVE SUBJECTS IS SATISFIED. THERE IS ALLOWED TO BE AN UNCERTAINTY IN TRACEABILITY BECAUSE OF ELAPSED TIME BETWEEN CHANGES IN THE GROUP MEMBERSHIP AND THE ENFORCEMENT IN THE ACCESS CONTROL MECHANISMS. In addition, there is no need in a distributed processing system like a network to reauthen- ticate a user at each point in the network where a projec- tion of a user (via the subject operating on behalf of the user) into another remote subject takes place. _________________________ = Department of Defense Password Management Guide- __________ __ _______ ________ __________ _____ line, CSC-STD-002-85 ____ The passing of identifiers and/or authentication infor- mation from one component to another is usually done in sup- port to the implementation of the discretionary access con- trol (DAC). This support relates directly to the DAC regarding access by a user to a storage object in a dif- ferent NTCB partition than the one where the user was authenticated. Employing a forwarded identification implies additional reliance on the source and components along the path. 2.2.2.2 Audit + Statement from DoD 5200.28-STD THE TCB SHALL BE ABLE TO CREATE, MAINTAIN, AND PROTECT FROM MODIFICATION OR UNAUTHORIZED ACCESS OR DESTRUCTION AN AUDIT TRAIL OF ACCESSES TO THE OBJECTS IT PROTECTS. THE AUDIT DATA SHALL BE PROTECTED BY THE TCB SO THAT READ ACCESS TO IT IS LIMITED TO THOSE WHO ARE AUTHORIZED FOR AUDIT DATA. THE TCB SHALL BE ABLE TO RECORD THE FOLLOWING TYPES OF EVENTS: USE OF IDENTIFICATION AND AUTHENTICATION MECHANISMS, INTRO- DUCTION OF OBJECTS INTO A USER'S ADDRESS SPACE (E.G., FILE OPEN, PROGRAM INITIATION), DELETION OF OBJECTS, ACTIONS TAKEN BY COMPUTER OPERATORS AND SYSTEM ADMINISTRATORS AND/OR SYSTEM SECURITY OFFICERS, AND OTHER SECURITY RELEVANT EVENTS. FOR EACH RECORDED EVENT, THE AUDIT RECORD SHALL IDENTIFY: DATE AND TIME OF THE EVENT, USER, TYPE OF EVENT, AND SUCCESS OR FAILURE OF THE EVENT. FOR IDENTIFICATION/AUTHENTICATION EVENTS THE ORIGIN OF REQUEST (E.G., TERMINAL ID) SHALL BE INCLUDED IN THE AUDIT RECORD. FOR EVENTS THAT INTRODUCE AN OBJECT INTO A USER'S ADDRESS SPACE AND FOR OBJECT DELETION EVENTS THE AUDIT RECORD SHALL INCLUDE THE NAME OF THE OBJECT. THE ADP SYSTEM ADMINISTRA- TOR SHALL BE ABLE TO SELECTIVELY AUDIT THE ACTIONS OF ANY ONE OR MORE USERS BASED ON INDIVIDUAL IDENTITY. + Interpretation THIS CRITERION APPLIES AS STATED. THE SPONSOR MUST SELECT WHICH EVENTS ARE AUDITABLE. IF ANY SUCH EVENTS ARE NOT DISTINGUISHABLE BY THE NTCB ALONE (FOR EXAMPLE THOSE IDENTIFIED IN PART II), THE AUDIT MECHANISM SHALL PROVIDE AN INTERFACE, WHICH AN AUTHORIZED SUBJECT CAN INVOKE WITH PARAMETERS SUFFICIENT TO PRODUCE AN AUDIT RECORD. THESE AUDIT RECORDS SHALL BE DISTINGUISHABLE FROM THOSE PROVIDED BY THE NTCB. IN THE CONTEXT OF A NETWORK SYSTEM, "OTHER SECURITY RELEVANT EVENTS" (DEPENDING ON NETWORK SYSTEM ARCHITECTURE AND NETWORK SECURITY POLICY) MIGHT BE AS FOL- LOWS: 1. IDENTIFICATION OF EACH ACCESS EVENT (E.G., ESTAB- LISHING A CONNECTION OR A CONNECTIONLESS ASSOCIATION BETWEEN PROCESSES IN TWO HOSTS OF THE NETWORK) AND ITS PRINCIPAL PARAMETERS (E.G., HOST IDENTIFIERS OF THE TWO HOSTS INVOLVED IN THE ACCESS EVENT AND USER IDENTIFIER OR HOST IDENTIFIER OF THE USER OR HOST THAT IS REQUESTING THE ACCESS EVENT) 2. IDENTIFICATION OF THE STARTING AND ENDING TIMES OF EACH ACCESS EVENT USING LOCAL TIME OR GLOBAL SYN- CHRONIZED TIME 3. IDENTIFICATION OF SECURITY-RELEVANT EXCEPTIONAL CON- DITIONS (E.G., POTENTIAL VIOLATION OF DATA INTEGRITY, SUCH AS MISROUTED DATAGRAMS) DETECTED DURING THE TRANSACTIONS BETWEEN TWO HOSTS 4. UTILIZATION OF CRYPTOGRAPHIC VARIABLES 5. CHANGING THE CONFIGURATION OF THE NETWORK (E.G., A COMPONENT LEAVING THE NETWORK AND REJOINING) IN ADDITION, IDENTIFICATION INFORMATION SHOULD BE INCLUDED IN APPROPRIATE AUDIT TRAIL RECORDS, AS NECESSARY, TO ALLOW ASSOCIATION OF ALL RELATED (E.G., INVOLVING THE SAME NETWORK EVENT) AUDIT TRAIL RECORDS (E.G., AT DIFFERENT HOSTS) WITH EACH OTHER. FURTHERMORE, A COMPONENT OF THE NETWORK SYSTEM MAY PROVIDE THE REQUIRED AUDIT CAPABILITY (E.G., STORAGE, RETRIEVAL, REDUCTION, ANALYSIS) FOR OTHER COMPONENTS THAT DO NOT INTERNALLY STORE AUDIT DATA BUT TRANSMIT THE AUDIT DATA TO SOME DESIGNATED COLLECTION COM- PONENT. PROVISIONS SHALL BE MADE TO CONTROL THE LOSS OF AUDIT DATA DUE TO UNAVAILABILITY OF RESOURCES. IN THE CONTEXT OF A NETWORK SYSTEM, THE "USER'S ADDRESS SPACE" IS EXTENDED, FOR OBJECT INTRODUCTION AND DELETION EVENTS, TO INCLUDE ADDRESS SPACES BEING EMPLOYED ON BEHALF OF A REMOTE USER (OR HOST). HOWEVER, THE FOCUS REMAINS ON USERS IN CONTRAST TO INTERNAL SUBJECTS AS DISCUSSED IN THE DAC CRITERION. IN ADDITION, AUDIT INFORMATION MUST BE STORED IN MACHINE-READABLE FORM. + Rationale FOR REMOTE USERS, THE NETWORK IDENTIFIERS (E.G., INTER- NET ADDRESS) CAN BE USED AS IDENTIFIERS OF GROUPS OF INDIVI- DUAL USERS (E.G., "ALL USERS AT HOST A") TO ELIMINATE THE MAINTENANCE THAT WOULD BE REQUIRED IF INDIVIDUAL IDENTIFICA- TION OF REMOTE USERS WAS EMPLOYED. IN THIS CLASS (C2), HOW- EVER, IT MUST BE POSSIBLE TO IDENTIFY (IMMEDIATELY OR AT SOME LATER TIME) THE INDIVIDUALS REPRESENTED BY A GROUP IDENTIFIER. IN ALL OTHER RESPECTS, THE INTERPRETATION IS A STRAIGHTFORWARD EXTENSION OF THE CRITERION INTO THE CONTEXT OF A NETWORK SYSTEM. 2.2.3 Assurance _ _ _ _________ 2.2.3.1 Operational Assurance 2.2.3.1.1 System Architecture + Statement from DoD 5200.28-STD The TCB shall maintain a domain for its own execution that protects it from external interference or tampering (e.g., by modification of its code or data structures). Resources controlled by the TCB may be a defined subset of the sub- jects and objects in the ADP system. THE TCB SHALL ISOLATE THE RESOURCES TO BE PROTECTED SO THAT THEY ARE SUBJECT TO THE ACCESS CONTROL AND AUDITING REQUIREMENTS. + Interpretation The system architecture criterion must be met individu- ally by all NTCB partitions. Implementation of the require- ment that the NTCB maintain a domain for its own execution is achieved by having each NTCB partition maintain a domain for its own execution. The subset of network resources over which the NTCB has control are the union of the sets of resources over which the NTCB partitions have control. Code and data structures belonging to the NTCB, transferred among NTCB subjects (i.e., subjects outside the reference monitor but inside the NTCB) belonging to different NTCB partitions, must be pro- tected against external interference or tampering. For example, a cryptographic checksum or physical means may be employed to protect user authentication data exchanged between NTCB partitions. EACH NTCB PARTITION PROVIDES ISOLATION OF RESOURCES (WITHIN ITS COMPONENT) TO BE PROTECTED IN ACCORD WITH THE NETWORK SYSTEM ARCHITECTURE AND SECURITY POLICY. + Rationale The requirement for the protection of communications between NTCB partitions is specifically directed to subjects that are part of the NTCB partitions. Any requirements for such protection for the subjects that are outside the NTCB partitions are addressed in response to the integrity requirements of the security policy. ISOLATION OF THE RESOURCES TO BE PROTECTED PROVIDES ADDITIONAL PROTECTION, COMPARED TO CLASS (C1), THAT MECHAN- ISMS THAT DEPEND ON THE RESOURCE (E.G., DAC AND USER IDEN- TIFICATION) WILL OPERATE CORRECTLY. 2.2.3.1.2 System Integrity + Statement from DoD 5200.28-STD Hardware and/or software features shall be provided that can be used to periodically validate the correct operation of the on-site hardware and firmware elements of the TCB. + Interpretation Implementation of the requirement is partly achieved by having hardware and/or software features that can be used to periodically validate the correct operation of the hardware and firmware elements of each component's NTCB partition. Features shall also be provided to validate the identity and correct operation of a component prior to its incorporation in the network system and throughout system operation. For example, a protocol could be designed that enables the com- ponents of the partitioned NTCB to exchange messages period- ically and validate each other's correct response. The pro- tocol shall be able to determine the remote entity's ability to respond. NTCB partitions shall provide the capability to report to network administrative personnel the failures detected in other NTCB partitions. Intercomponent protocols implemented within a NTCB shall be designed in such a way as to provide correct opera- tion in the case of failures of network communications or individual components. The allocation of discretionary access control policy in a network may require communication between trusted subjects that are part of the NTCB parti- tions in different components. This communication is nor- mally implemented with a protocol between the subjects as peer entities. Incorrect access within a component shall not result from failure of an NTCB partition to communicate with other components. + Rationale The first paragraph of the interpretation is a straightforward extension of the requirement into the con- text of a network system and partitioned NTCB as defined for these network criteria. NTCB protocols should be robust enough so that they permit the system to operate correctly in the case of local- ized failure. The purpose of this protection is to preserve the integrity of the NTCB itself. It is not unusual for one or more components in a network to be inoperative at any time, so it is important to minimize the effects of such failures on the rest of the network. Additional integrity and denial of service issues are addressed in Part II. 2.2.3.2 Life-Cycle Assurance 2.2.3.2.1 Security Testing + Statement from DoD 5200.28-STD The security mechanisms of the ADP system shall be tested and found to work as claimed in the system documentation. Testing shall be done to assure that there are no obvious ways for an unauthorized user to bypass or otherwise defeat the security protection mechanisms of the TCB. TESTING SHALL ALSO INCLUDE A SEARCH FOR OBVIOUS FLAWS THAT WOULD ALLOW VIOLATION OF RESOURCE ISOLATION, OR THAT WOULD PERMIT UNAU- THORIZED ACCESS TO THE AUDIT OR AUTHENTICATION DATA. (See the Security Testing Guidelines.) + Interpretation Testing of a component will require a testbed that exercises the interfaces and protocols of the COMPONENT INCLUDING TESTS UNDER EXCEPTIONAL CONDITIONS. The testing of a security mechanism of the network system for meeting this criterion shall be an integrated testing procedure involving all components containing an NTCB partition that implement the given mechanism. This integrated testing is additional to any individual component tests involved in the evaluation of the network system. The sponsor should iden- tify the allowable set of configurations including the sizes of the networks. Analysis or testing procedures and tools shall be available to test the limits of these configura- tions. A change in configuration within the allowable set of configurations does not require retesting. + Rationale Testing is the primary method available in this evalua- tion division to gain any assurance that the security mechanisms perform their intended function. 2.2.4 Documentation _ _ _ _____________ 2.2.4.1 Security Features User's Guide + Statement from DoD 5200.28-STD A single summary, chapter, or manual in user documentation shall describe the protection mechanisms provided by the TCB, interpretations on their use, and how they interact with one another. + Interpretation This user documentation describes user visible protec- tion mechanisms at the global (network system) level and at the user interface of each component, and the interaction among these. + Rationale The interpretation is an extension of the requirement into the context of a network system as defined for these network criteria. Documentation of protection mechanisms provided by individual components is required by the cri- teria for trusted computer systems that are applied as appropriate for the individual components. 2.2.4.2 Trusted Facility Manual + Statement from DoD 5200.28-STD A manual addressed to the ADP system administrator shall present cautions about functions and privileges that should be controlled when running a secure facility. THE PROCEDURES FOR EXAMINING AND MAINTAINING THE AUDIT FILES AS WELL AS THE DETAILED AUDIT RECORD STRUCTURE FOR EACH TYPE OF AUDIT EVENT SHALL BE GIVEN. + Interpretation This manual shall contain specifications and procedures to assist the system administrator(s) maintain cognizance of the network configuration. These specifications and pro- cedures shall address the following: 1. The hardware configuration of the network itself; 2. The implications of attaching new components to the network; 3. The case where certain components may periodically leave the network (e.g., by crashing, or by being disconnected) and then rejoin; 4. Network configuration aspects that can impact the security of the network system; (For example, the manual should describe for the network system administrator the interconnections among components that are consistent with the overall network system architecture.) 5. Loading or modifying NTCB software or firmware (e.g., down-line loading). The physical and administrative environmental controls shall be specified. Any assumptions about security of a given network should be clearly stated (e.g., the fact that all communications links must be physically protected to a certain level). + Rationale There may be multiple system administrators with diverse responsibilities. The technical security measures described by these criteria must be used in conjunction with other forms of security in order to achieve security of the network. Additional forms include administrative security, physical security, emanations security, etc. Extension of this criterion to cover configuration aspects of the network is needed because, for example, proper interconnection of components is typically essential to achieve a correct realization of the network architec- ture. Cryptography is one common mechanism employed to pro- tect communication circuits. Encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity of the ciphertext is generally lower than the cleartext. If encryption methodologies are employed, they shall be approved by the National Security Agency (NSA). The encryption algorithm and its implementation are outside the scope of these interpretations. This algorithm and implementation may be implemented in a separate device or may be a function of a subject in a component not dedi- cated to encryption. Without prejudice, either implementa- tion packaging is referred to as an encryption mechanism herein. 2.2.4.3 Test Documentation + Statement from DoD 5200.28-STD The system developer shall provide to the evaluators a docu- ment that describes the test plan, test procedures that show how the security mechanisms were tested, and results of the security mechanisms' functional testing. + Interpretation The "system developer" is interpreted as "the network system sponsor". The description of the test plan should establish the context in which the testing was or should be conducted. The description should identify any additional test components that are not part of the system being evaluated. This includes a description of the test-relevant functions of such test components and a description of the interfacing of those test components to the system being evaluated. The description of the test plan should also demonstrate that the tests adequately cover the network security policy. The tests should include the features described in the System Architecture and the System Integrity sections. The tests should also include network configuration and sizing. + Rationale The entity being evaluated may be a networking subsys- tem (see Appendix A) to which other components must be added to make a complete network system. In that case, this interpretation is extended to include contextual definition because, at evaluation time, it is not possible to validate the test plans without the description of the context for testing the networking subsystem. 2.2.4.4 Design Documentation + Statement from DoD 5200.28-STD Documentation shall be available that provides a description of the manufacturer's philosophy of protection and an expla- nation of how this philosophy is translated into the TCB. If the TCB is composed of distinct modules, the interfaces between these modules shall be described. + Interpretation Explanation of how the sponsor's philosophy of protec- tion is translated into the NTCB shall include a description of how the NTCB is partitioned. The security policy also shall be stated. The description of the interfaces between the NTCB modules shall include the interface(s) between NTCB partitions and modules within the partitions if the modules exist. The sponsor shall describe the security architecture and design, including the allocation of security require- ments among components. Appendix A addresses component evaluation issues. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system as defined for this network interpretation. Other documenta- tion, such as description of components and description of operating environment(s) in which the networking subsystem or network system is designed to function, is required else- where, e.g., in the Trusted Facility Manual. In order to be evaluated, a network must possess a coherent Network Security Architecture and Design. (Inter- connection of components that do not adhere to such a single coherent Network Security Architecture is addressed in the Interconnection of Accredited AIS, Appendix C.) The Network Security Architecture must address the security-relevant policies, objectives, and protocols. The Network Security Design specifies the interfaces and services that must be incorporated into the network so that it can be evaluated as a trusted entity. There may be multiple designs that con- form to the same architecture but are more or less incompa- tible and non-interoperable (except through the Interconnec- tion Rules). Security related mechanisms requiring coopera- tion among components are specified in the design in terms of their visible interfaces; mechanisms having no visible interfaces are not specified in this document but are left as implementation decisions. The Network Security Architecture and Design must be available from the network sponsor before evaluation of the network, or any component, can be undertaken. The Network Security Architecture and Design must be sufficiently com- plete, unambiguous, and free from obvious flaws to permit the construction or assembly of a trusted network based on the structure it specifies. When a component is being designed or presented for evaluation, or when a network assembled from components is assembled or presented for evaluation, there must be a priori evidence that the Network security Architecture and Design are satisfied. That is, the components can be assem- bled into a network that conforms in every way with the Net- work Security Architecture and Design to produce a physical realization that is trusted to the extent that its evalua- tion indicates. In order for a trusted network to be constructed from components that can be built independently, the Network Security Architecture and Design must completely and unambi- guously define the security functionality of components as well as the interfaces between or among components. The Network Security Architecture and Design must be evaluated to determine that a network constructed to its specifica- tions will in fact be trusted, that is, it will be evaluat- able under these interpretations. 3.0 DIVISION B: MANDATORY PROTECTION The notion of an NTCB that preserves the integrity of sensi- tivity labels and uses them to enforce a set of mandatory access control rules is a major requirement in this divi- sion. Network systems in this division must carry the sen- sitivity labels with major data structures in the system. The network system sponsor also provides the security policy model on which the NTCB is based and furnishes a specifica- tion of the NTCB. Evidence must be provided to demonstrate that the reference monitor concept has been implemented. 3.1 CLASS (B1): LABELED SECURITY PROTECTION _ _ _____ __ _______ ________ __________ CLASS (B1) NETWORK SYSTEMS REQUIRE ALL THE FEATURES REQUIRED FOR CLASS (C2). IN ADDITION, AN INFORMAL STATEMENT OF THE SECURITY POLICY MODEL, DATA LABELING, AND MANDATORY ACCESS CONTROL OVER SUBJECTS AND STORAGE OBJECTS MUST BE PRESENT. THE CAPABILITY MUST EXIST FOR ACCURATELY LABELING EXPORTED INFORMATION. ANY FLAWS IDENTIFIED BY TESTING MUST BE REMOVED. THE FOLLOWING ARE MINIMAL REQUIREMENTS FOR NETWORK SYSTEMS ASSIGNED A CLASS (B1) RATING: 3.1.1 Security Policy _ _ _ ________ ______ + Statement from DoD 5200.28-STD Implied from the Introduction to the TCSEC. + Interpretation The network sponsor shall describe the overall network security policy enforced by the NTCB. At a minimum, this policy shall include the discretionary AND MANDATORY requirements applicable to this class. The policy may require data secrecy, or data integrity, or both. THE POL- ICY IS AN ACCESS CONTROL POLICY HAVING TWO PRIMARY COM- PONENTS: MANDATORY AND DISCRETIONARY. The policy shall include a discretionary policy for protecting the informa- tion being processed based on the authorizations of indivi- duals, users, or groups of users. This access control pol- icy statement shall describe the requirements on the network to prevent or detect "reading or destroying" sensitive information by unauthorized users or errors. THE MANDATORY POLICY MUST DEFINE THE SET OF DISTINCT SENSITIVITY LEVELS THAT IT SUPPORTS. FOR THE CLASS B1 OR ABOVE THE MANDATORY POLICY SHALL BE BASED ON THE LABELS ASSOCIATED WITH THE INFORMATION THAT REFLECTS ITS SENSITIVITY WITH RESPECT TO SECRECY AND/OR INTEGRITY, WHERE APPLICABLE, AND LABELS ASSO- CIATED WITH USERS TO REFLECT THEIR AUTHORIZATION TO ACCESS SUCH INFORMATION. UNAUTHORIZED USERS INCLUDE BOTH THOSE that are not authorized to use the network at all (e.g., a user attempting to use a passive or active wire tap) or a legitimate user of the network who is not authorized to access a specific piece of information being protected. Note that "users" does not include "operators," "system programmers," "technical control officers," "system security officers," and other system support personnel. They are distinct from users and are subject to the Trusted Facility Manual and the System Architecture requirements. Such indi- viduals may change the system parameters of the network sys- tem, for example, by defining membership of a group. These individuals may also have the separate role of users. SECRECY POLICY: The network sponsor shall define the form of the discretionary AND MANDATORY secrecy policy that is enforced in the network to prevent unauthorized users from reading the sensitive infor- mation entrusted to the network. DATA INTEGRITY POLICY: The network sponsor shall define the discretionary AND MANDATORY integrity policy to prevent unauthorized users from modifying, viz., writing, sensitive information. The defini- tion of data integrity presented by the network sponsor refers to the requirement that the informa- tion has not been subjected to unauthorized modifi- cation in the network. THE MANDATORY INTEGRITY POL- ICY ENFORCED BY THE NTCB CANNOT, IN GENERAL, PREVENT MODIFICATION WHILE INFORMATION IS BEING TRANSMITTED BETWEEN COMPONENTS. HOWEVER, AN INTEGRITY SENSI- TIVITY LABEL MAY REFLECT THE CONFIDENCE THAT THE INFORMATION HAS NOT BEEN SUBJECTED TO TRANSMISSION ERRORS BECAUSE OF THE PROTECTION AFFORDED DURING TRANSMISSION. THIS REQUIREMENT IS DISTINCT FROM THE REQUIREMENT FOR LABEL INTEGRITY. + Rationale The word "sponsor" is used in place of alternatives (such as "vendor," "architect," "manufacturer," and "developer") because the alternatives indicate people who may not be available, involved, or relevant at the time that a network system is proposed for evaluation. A trusted network is able to control both the reading and writing of shared sensitive information. Control of writing is used to protect against destruction of informa- tion. A network normally is expected to have policy require- ments to protect both the secrecy and integrity of the information entrusted to it. In a network the integrity is frequently as important or more important than the secrecy requirements. Therefore the secrecy and/or integrity policy to be enforced by the network must be stated for each net- work regardless of its evaluation class. The assurance that the policy is faithfully enforced is reflected in the evaluation class of the network. This control over modification is typically used to protect information so that it may be relied upon and to control the potential harm that would result if the informa- tion were corrupted. The overall network policy require- ments for integrity includes the protection for data both while being processed in a component and while being transmitted in the network. The access control policy enforced by the NTCB relates to the access of subjects to objects within each component. Communications integrity addressed within Part II relates to information while being transmitted. THE MANDATORY INTEGRITY POLICY (AT CLASS B1 AND ABOVE) IN SOME ARCHITECTURES MAY BE USEFUL IN SUPPORTING THE LINK- AGE BETWEEN THE CONNECTION ORIENTED ABSTRACTION INTRODUCED IN THE INTRODUCTION AND THE INDIVIDUAL COMPONENTS OF THE NETWORK. FOR EXAMPLE, IN A KEY DISTRIBUTION CENTER FOR END-TO-END ENCRYPTION, A DISTINCT INTEGRITY CATEGORY MAY BE ASSIGNED TO ISOLATE THE KEY GENERATION CODE AND DATA FROM POSSIBLE MODIFICATION BY OTHER SUPPORTING PROCESSES IN THE SAME COMPONENT, SUCH AS OPERATOR INTERFACES AND AUDIT. THE MANDATORY INTEGRITY POLICY FOR SOME ARCHITECTURE MAY DEFINE AN INTEGRITY SENSITIVITY LABEL THAT REFLECTS THE SPECIFIC REQUIREMENTS FOR ENSURING THAT INFORMATION HAS NOT BEEN SUBJECT TO RANDOM ERRORS IN EXCESS OF A STATED LIMIT NOR TO UNAUTHORIZED MESSAGE STREAM MODIFICATION (MSM) -. THE SPECIFIC METRIC ASSOCIATED WITH AN INTEGRITY SENSITIVITY LABEL WILL GENERALLY REFLECT THE INTENDED APPLICATIONS OF THE NETWORK. 3.1.1.1 Discretionary Access Control + Statement from DoD 5200.28-STD The TCB shall define and control access between named users and named objects (e.g., files and programs) in the ADP sys- tem. The enforcement mechanism (e.g., self/group/public controls, access control lists) shall allow users to specify and control sharing of those objects by named individuals or defined groups of individuals, or both, and shall provide controls to limit propagation of access rights. The discre- tionary access control mechanism shall, either by explicit user action or by default, provide that objects are pro- tected from unauthorized access. These access controls shall be capable of including or excluding access to the granularity of a single user. Access permission to an object by users not already possessing access permission shall only be assigned by authorized users. _________________________ - See Voydock, Victor L. and Stephen T. Kent, "Secu- rity Mechanisms in High-Level Network Protocols," Com- ___ puting Surveys, Vol. 15, No. 2, June 1983, pp 135-171. ______ _______ + Interpretation The discretionary access control (DAC) mechanism(s) may be distributed over the partitioned NTCB in various ways. Some part, all, or none of the DAC may be implemented in a given component of the network system. In particular, com- ponents that support only internal subjects (i.e., that have no subjects acting as direct surrogates for users), such as a public network packet switch, might not implement the DAC mechanism(s) directly (e.g., they are unlikely to contain access control lists). Identification of users by groups may be achieved in various ways in the networking environment. For example, the network identifiers (e.g., internet addresses) for vari- ous components (e.g., hosts, gateways) can be used as iden- tifiers of groups of individual users (e.g., "all users at Host A," "all users of network Q") so long as the individu- als involved in the group are implied by the group identif- ier. For example, Host A might employ a particular group-id, for which it maintains a list of explicit users in that group, in its network exchange with Host B, which accepts the group-id under the conditions of this interpretation. For networks, individual hosts will impose need-to-know controls over their users on the basis of named individuals - much like (in fact, probably the same) controls used when there is no network connection. When group identifiers are acceptable for access con- trol, the identifier of some other host may be employed, to eliminate the maintenance that would be required if indivi- dual identification of remote users was employed. In class C2 and higher, however, it must be possible from that audit record to identify (immediately or at some later time) exactly the individuals represented by a group identifier at the time of the use of that identifier. There is allowed to be an uncertainty because of elapsed time between changes in the group membership and the enforcement in the access con- trol mechanisms. The DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. The reference monitor manages all the physical resources of the system and from them creates the abstraction of subjects and objects that it con- trols. Some of these subjects and objects may be used to implement a part of the NTCB. When the DAC mechanism is distributed in such NTCB subjects (i.e., when outside the reference monitor), the assurance requirements (see the Assurance section) for the design and implementation of the DAC shall be those of class C2 for all networks of class C2 or above. When integrity is included as part of the network dis- cretionary security policy, the above interpretations shall be specifically applied to the controls over modification, viz, the write mode of access, within each component based on identified users or groups of users. + Rationale In this class, the supporting elements of the overall DAC mechanism are required to isolate information (objects) that supports DAC so that it is subject to auditing require- ments (see the System Architecture section). The use of network identifiers to identify groups of individual users could be implemented, for example, as an X.25 community of interest in the network protocol layer (layer 3). In all other respects, the supporting elements of the overall DAC mechanism are treated exactly as untrusted subjects are treated with respect to DAC in an ADP system, with the same result as noted in the interpretation. A typical situation for DAC is that a surrogate process for a remote user will be created in some host for access to objects under the control of the NTCB partition within that host. The interpretation requires that a user identifier be assigned and maintained for each such process by the NTCB, so that access by a surrogate process is subject to essen- tially the same discretionary controls as access by a pro- cess acting on behalf of a local user would be. However, within this interpretation a range of possible interpreta- tions of the assigned user identification is permitted. The most obvious situation would exist if a global database of network users were to be made permanently avail- able on demand to every host, (i.e., a name server existed) so that all user identifications were globally meaningful. It is also acceptable, however, for some NTCB parti- tions to maintain a database of locally-registered users for its own use. In such a case, one could choose to inhibit the creation of surrogate processes for locally unregistered users, or (if permitted by the local policy) alternatively, to permit the creation of surrogate processes with preselected user and group identifiers which, in effect, identify the process as executing on behalf of a member of a group of users on a particular remote host. The intent of the words concerning audit in the interpretation is to pro- vide a minimally acceptable degree of auditability for cases such as the last described. What is required is that there be a capability, using the audit facilities provided by the network NTCB partitions involved, to determine who was logged in at the actual host of the group of remote users at the time the surrogate processing occured. Associating the proper user id with a surrogate process is the job of identification and authentication. This means that DAC is applied locally, with respect to the user id of the surrogate process. The transmission of the data back across the network to the user's host, and the creation of a copy of the data there, is not the business of DAC. Components that support only internal subjects impact the implementation of the DAC by providing services by which information (e.g., a user-id) is made available to a com- ponent that makes a DAC decision. An example of the latter would be the case that a user at Host A attempts to access a file at Host B. The DAC decision might be (and usually would be) made at Host B on the basis of a user-id transmit- ted from Host A to Host B. Unique user identification may be achieved by a variety of mechanisms, including (a) a requirement for unique iden- tification and authentication on the host where access takes place; (b) recognition of fully qualified network addresses authenticated by another host and forwarded to the host where access takes place; or (c) administrative support of a network-wide unique personnel identifier that could be authenticated and forwarded by another host as in (b) above, or could be authenticated and forwarded by a dedicated net- work identification and authentication server. The proto- cols which implement (b) or (c) are subject to the System Architecture requirements. Network support for DAC might be handled in other ways than that described as "typical" above. In particular, some form of centralized access control is often proposed. An access control center may make all decisions for DAC, or it may share the burden with the hosts by controlling host-to- host connections, and leaving the hosts to decide on access to their objects by users at a limited set of remote hosts. In this case the access control center provides the linkage between the connection oriented abstraction (as discussed in the Introduction) and the overall network security policy for DAC. In all cases the enforcement of the decision must be provided by the host where the object resides. THERE ARE TWO FORMS OF DISTRIBUTION FOR THE DAC MECHAN- ISM: IMPLEMENTING PORTIONS OF THE DAC IN SEPARATE COM- PONENTS, AND SUPPORTING THE DAC IN SUBJECTS CONTAINED WITHIN THE NTCB PARTITION IN A COMPONENT. SINCE "THE ADP SYSTEM" IS UNDERSTOOD TO BE "THE COMPUTER NETWORK" AS A WHOLE, EACH NETWORK COMPONENT IS RESPONSIBLE FOR ENFORCING SECURITY IN THE MECHANISMS ALLOCATED TO IT TO ENSURE SECURE IMPLEMENTA- TION OF THE NETWORK SECURITY POLICY. FOR TRADITIONAL HOST SYSTEMS IT IS FREQUENTLY EASY TO ALSO ENFORCE THE DAC ALONG WITH THE MAC WITHIN THE REFERENCE MONITOR, PER SE, ALTHOUGH A FEW APPROACHES, SUCH AS VIRTUAL MACHINE MONITORS, SUPPORT DAC OUTSIDE THIS INTERFACE. IN CONTRAST TO THE UNIVERSALLY RIGID STRUCTURE OF MAN- DATORY POLICIES (SEE THE MANDATORY ACCESS CONTROL SECTION), DAC POLICIES TEND TO BE VERY NETWORK AND SYSTEM SPECIFIC, WITH FEATURES THAT REFLECT THE NATURAL USE OF THE SYSTEM. FOR NETWORKS IT IS COMMON THAT INDIVIDUAL HOSTS WILL IMPOSE CONTROLS OVER THEIR LOCAL USERS ON THE BASIS OF NAMED INDIVIDUALS-MUCH LIKE THE CONTROLS USED WHEN THERE IS NO NETWORK CONNECTION. HOWEVER, IT IS DIFFICULT TO MANAGE IN A CENTRALIZED MANNER ALL THE INDIVIDUALS USING A LARGE NET- WORK. THEREFORE, USERS ON OTHER HOSTS ARE COMMONLY GROUPED TOGETHER SO THAT THE CONTROLS REQUIRED BY THE NETWORK DAC POLICY ARE ACTUALLY BASED ON THE IDENTITY OF THE HOSTS OR OTHER COMPONENTS. A GATEWAY IS AN EXAMPLE OF SUCH A COM- PONENT. THE ASSURANCE REQUIREMENTS ARE AT THE VERY HEART OF THE CONCEPT OF A TRUSTED SYSTEM. IT IS THE ASSURANCE THAT DETERMINES IF A SYSTEM OR NETWORK IS APPROPRIATE FOR A GIVEN ENVIRONMENT, AS REFLECTED, FOR EXAMPLE, IN THE ENVIRONMENTS GUIDELINE-. IN THE CASE OF MONOLITHIC SYSTEMS THAT HAVE DAC INTEGRAL TO THE REFERENCE MONITOR, THE ASSURANCE REQUIRE- MENTS FOR DAC ARE INSEPARABLE FROM THOSE OF THE REST OF THE REFERENCE MONITOR. FOR NETWORKS THERE IS TYPICALLY A MUCH CLEARER DISTINCTION DUE TO DISTRIBUTED DAC. THE RATIONALE FOR MAKING THE DISTINCTION IN THIS NETWORK INTERPRETATION IS THAT IF MAJOR TRUSTED NETWORK COMPONENTS CAN BE MADE SIGNI- FICANTLY EASIER TO DESIGN AND IMPLEMENT WITHOUT REDUCING THE ABILITY TO MEET SECURITY POLICY, THEN TRUSTED NETWORKS WILL BE MORE EASILY AVAILABLE. 3.1.1.2 Object Reuse + Statement from DoD 5200.28-STD All authorizations to the information contained within a storage object shall be revoked prior to initial assignment, allocation or reallocation to a subject from the TCB's pool of unused storage objects. No information, including encrypted representations of information, produced by a prior subject's actions is to be available to any subject that obtains access to an object that has been released back to the system. + Interpretation The NTCB shall ensure that any storage objects that it controls (e.g., message buffers under the control of a NTCB partition in a component) contain no information for which a subject in that component is not authorized before granting access. This requirement must be enforced by each of the NTCB partitions. _________________________ - Guidance for Applying the Department of Defense ________ ___ ________ ___ __________ __ _______ Trusted Computer System Evaluation Criteria in Specific _______ ________ ______ __________ ________ __ ________ Environments, CSC-STD-003-85. ____________ + Rationale In a network system, storage objects of interest are things that the NTCB directly controls, such as message buffers in components. Each component of the network system must enforce the object reuse requirement with respect to the storage objects of interest as determined by the network security policy. For example, the DAC requirement in this division leads to the requirement here that message buffers be under the control of the NTCB partition. A buffer assigned to an internal subject may be reused at the discre- tion of that subject which is responsible for preserving the integrity of message streams. Such controlled objects may be implemented in physical resources, such as buffers, disk sectors, tape space, and main memory, in components such as network switches. 3.1.1.3 Labels + Statement from DoD 5200.28-STD SENSITIVITY LABELS ASSOCIATED WITH EACH SUBJECT AND STORAGE OBJECT UNDER ITS CONTROL (E.G., PROCESS, FILE, SEGMENT, DEV- ICE) SHALL BE MAINTAINED BY THE TCB. THESE LABELS SHALL BE USED AS THE BASIS FOR MANDATORY ACCESS CONTROL DECISIONS. IN ORDER TO IMPORT NON-LABELED DATA, THE TCB SHALL REQUEST AND RECEIVE FROM AN AUTHORIZED USER THE SENSITIVITY LEVEL OF THE DATA, AND ALL SUCH ACTIONS SHALL BE AUDITABLE BY THE TCB. + Interpretation NON-LABELED DATA IMPORTED UNDER THE CONTROL OF THE NTCB PARTITION WILL BE ASSIGNED A LABEL CONSTRAINED BY THE SINGLE-LEVEL DEVICE USED TO IMPORT IT. LABELS MAY INCLUDE SECRECY AND INTEGRITY- COMPONENTS IN ACCORDANCE WITH THE OVERALL NETWORK SECURITY POLICY DESCRIBED BY THE NETWORK SPONSOR. WHENEVER THE TERM "LABEL" IS USED THROUGHOUT THIS INTERPRETATION, IT IS UNDERSTOOD TO INCLUDE BOTH COMPONENTS AS APPLICABLE. SIMILARLY, THE TERMS "SINGLE-LEVEL" AND "MULTILEVEL" ARE UNDERSTOOD TO BE BASED ON BOTH THE SECRECY AND INTEGRITY COMPONENTS OF THE POLICY. THE MANDATORY INTEGRITY POLICY WILL TYPICALLY HAVE REQUIREMENTS, SUCH AS THE PROBABILITY OF UNDETECTED MESSAGE STREAM MODIFICATION, THAT WILL BE REFLECTED IN THE LABEL FOR THE DATA SO PRO- TECTED. FOR EXAMPLE, WHEN DATA IS IMPORTED ITS INTEGRITY LABEL MAY BE ASSIGNED BASED ON MECHANISMS, SUCH AS CRYPTOG- RAPHY, USED TO PROVIDE THE ASSURANCE REQUIRED BY THE POLICY. THE NTCB SHALL ASSURE THAT SUCH MECHANISM ARE PROTECTED FROM TAMPERING AND ARE ALWAYS INVOKED WHEN THEY ARE THE BASIS FOR _________________________ - See, for example, Biba, K.J., "Integrity Considera- tion for Secure Computer Systems," ESD-TR-76-372, MTR- 3153, The MITRE Corporation, Bedford, MA, April 1977. A LABEL. + Rationale THE INTERPRETATION IS AN EXTENSION OF THE REQUIREMENT INTO THE CONTEXT OF A NETWORK SYSTEM AND PARTITIONED NTCB AS DEFINED FOR THESE NETWORK INTERPRETATIONS. A SINGLE-LEVEL DEVICE MAY BE REGARDED EITHER AS A SUBJECT OR AN OBJECT. A MULTILEVEL DEVICE IS REGARDED AS A TRUSTED SUBJECT IN WHICH THE SECURITY RANGE OF THE SUBJECT IS THE MINIMUM-MAXIMUM RANGE OF THE DATA EXPECTED TO BE TRANSMITTED OVER THE DEV- ICE. THE SENSITIVITY LABELS FOR EITHER SECRECY OR INTEGRITY OR BOTH MAY REFLECT NON-HIERARCHICAL CATEGORIES OR HIERARCH- ICAL CLASSIFICATION OR BOTH. 3.1.1.3.1 Label Integrity + Statement from DoD 5200.28-STD SENSITIVITY LABELS SHALL ACCURATELY REPRESENT SENSITIVITY LEVELS OF THE SPECIFIC SUBJECTS OR OBJECTS WITH WHICH THEY ARE ASSOCIATED. WHEN EXPORTED BY THE TCB, SENSITIVITY LABELS SHALL ACCURATELY AND UNAMBIGUOUSLY REPRESENT THE INTERNAL LABELS AND SHALL BE ASSOCIATED WITH THE INFORMATION BEING EXPORTED. + Interpretation THE PHRASE "EXPORTED BY THE TCB" IS UNDERSTOOD TO INCLUDE TRANSMISSION OF INFORMATION FROM AN OBJECT IN ONE COMPONENT TO AN OBJECT IN ANOTHER COMPONENT. INFORMATION TRANSFERRED BETWEEN NTCB PARTITIONS IS ADDRESSED IN THE SYS- TEM INTEGRITY SECTION. THE FORM OF INTERNAL AND EXTERNAL (EXPORTED) SENSITIVITY LABELS MAY DIFFER, BUT THE MEANING SHALL BE THE SAME. THE NTCB SHALL, IN ADDITION, ENSURE THAT CORRECT ASSOCIATION OF SENSITIVITY LABELS WITH THE INFORMA- TION BEING TRANSPORTED ACROSS THE NETWORK IS PRESERVED. AS MENTIONED IN THE TRUSTED FACILITY MANUAL SECTION, ENCRYPTION TRANSFORMS THE REPRESENTATION OF INFORMATION SO THAT IT IS UNINTELLIGIBLE TO UNAUTHORIZED SUBJECTS. REFLECTING THIS TRANSFORMATION, THE SENSITIVITY LEVEL OF THE CIPHERTEXT IS GENERALLY LOWER THAN THE CLEARTEXT. IT FOL- LOWS THAT CLEARTEXT AND CIPHERTEXT ARE CONTAINED IN DIF- FERENT OBJECTS, EACH POSSESSING ITS OWN LABEL. THE LABEL OF THE CLEARTEXT MUST BE PRESERVED AND ASSOCIATED WITH THE CIPHERTEXT SO THAT IT CAN BE RESTORED WHEN THE CLEARTEXT IS SUBSEQUENTLY OBTAINED BY DECRYPTING THE CIPHERTEXT. IF THE CLEARTEXT IS ASSOCIATED WITH A SINGLE-LEVEL DEVICE, THE LABEL OF THAT CLEARTEXT MAY BE IMPLICIT. THE LABEL MAY ALSO BE IMPLICIT IN THE KEY. WHEN INFORMATION IS EXPORTED TO AN ENVIRONMENT WHERE IT IS SUBJECT TO DELIBERATE OR ACCIDENTAL MODIFICATION, THE TCB SHALL SUPPORT THE MEANS, SUCH AS CRYPTOGRAPHIC CHECKSUMS, TO ASSURE THE ACCURACY OF THE LABELS. WHEN THERE IS A MANDA- TORY INTEGRITY POLICY, THE POLICY WILL DEFINE THE MEANING OF INTEGRITY LABELS. + Rationale ENCRYPTION ALGORITHMS AND THEIR IMPLEMENTATION ARE OUT- SIDE THE SCOPE OF THESE INTERPRETATIONS. SUCH ALGORITHMS MAY BE IMPLEMENTED IN A SEPARATE DEVICE OR MAY BE INCOR- PORATED IN A SUBJECT OF A LARGER COMPONENT. WITHOUT PREJU- DICE, EITHER IMPLEMENTATION PACKAGING IS REFERRED TO AS AN ENCRYPTION MECHANISM HEREIN. IF ENCRYPTION METHODOLOGIES ARE EMPLOYED IN THIS REGARD, THEY SHALL BE APPROVED BY THE NATIONAL SECURITY AGENCY (NSA). THE ENCRYPTION PROCESS IS PART OF THE NETWORK TRUSTED COMPUTER BASE PARTITION IN THE COMPONENTS IN WHICH IT IS IMPLEMENTED. THE ENCRYPTION MECHANISM IS NOT NECESSARILY A MUL- TILEVEL DEVICE OR MULTILEVEL SUBJECT, AS THESE TERMS ARE USED IN THESE CRITERIA. THE PROCESS OF ENCRYPTION IS MUL- TILEVEL BY DEFINITION. THE CLEARTEXT AND CIPHERTEXT INTER- FACES CARRY INFORMATION OF DIFFERENT SENSITIVITY. AN ENCRYPTION MECHANISM DOES NOT PROCESS DATA IN THE SENSE OF PERFORMING LOGICAL OR ARITHMETIC OPERATIONS ON THAT DATA WITH THE INTENT OF PRODUCING NEW DATA. THE CLEARTEXT AND CIPHERTEXT INTERFACES ON THE ENCRYPTION MECHANISM MUST BE SEPARATELY IDENTIFIED AS BEING SINGLE-LEVEL OR MULTILEVEL. IF THE INTERFACE IS SINGLE-LEVEL, THEN THE SENSITIVITY OF THE DATA IS ESTABLISHED BY A TRUSTED INDIVIDUAL AND IMPLI- CITLY ASSOCIATED WITH THE INTERFACE; THE EXPORTATION TO SINGLE-LEVEL DEVICES CRITERION APPLIES. IF THE INTERFACE IS MULTILEVEL, THEN THE DATA MUST BE LABELED; THE EXPORTATION TO MULTILEVEL DEVICES CRITERION APPLIES. THE NETWORK ARCHITECT IS FREE TO SELECT AN ACCEPT- TABLE MECHANISM FOR ASSOCIATING A LABEL WITH AN OBJECT. WITH REFERENCE TO ENCRYPTED OBJECTS, THE FOLLOWING EXAMPLES ARE POSSIBLE: 1. INCLUDE A LABEL FIELD IN THE PROTOCOL DEFINITION OF THE OBJECT. 2. IMPLICITLY ASSOCIATE THE LABEL WITH THE OBJECT THROUGH THE ENCRYPTION KEY. THAT IS, THE ENCRYPTION KEY UNIQUELY IDENTIFIES A SENSITIVITY LEVEL. A SIN- GLE OR PRIVATE KEY MUST BE PROTECTED AT THE LEVEL OF THE DATA THAT IT ENCRYPTS. 3.1.1.3.2 Exportation of Labeled Information + Statement from DoD 5200.28-STD THE TCB SHALL DESIGNATE EACH COMMUNICATION CHANNEL AND I/O DEVICE AS EITHER SINGLE-LEVEL OR MULTILEVEL. ANY CHANGE IN THIS DESIGNATION SHALL BE DONE MANUALLY AND SHALL BE AUDIT- ABLE BY THE TCB. THE TCB SHALL MAINTAIN AND BE ABLE TO AUDIT ANY CHANGE IN THE SENSITIVITY LEVEL OR LEVELS ASSOCI- ATED WITH A COMMUNICATIONS CHANNEL OR I/O DEVICE. + Interpretation EACH COMMUNICATION CHANNEL AND NETWORK COMPONENT SHALL BE DESIGNATED AS EITHER SINGLE-LEVEL OR MULTILEVEL. ANY CHANGE IN THIS DESIGNATION SHALL BE DONE WITH THE COGNIZANCE AND APPROVAL OF THE ADMINISTRATOR OR SECURITY OFFICER IN CHARGE OF THE AFFECTED COMPONENTS AND THE ADMINISTRATOR OR SECURITY OFFICER IN CHARGE OF THE NTCB. THIS CHANGE SHALL BE AUDITABLE BY THE NETWORK. THE NTCB SHALL MAINTAIN AND BE ABLE TO AUDIT ANY CHANGE IN THE CURRENT SENSITIVITY LEVEL ASSOCIATED WITH THE DEVICE CONNECTED TO A SINGLE-LEVEL COM- MUNICATION CHANNEL OR THE RANGE ASSOCIATED WITH A MULTILEVEL COMMUNICATION CHANNEL OR COMPONENT. THE NTCB SHALL ALSO BE ABLE TO AUDIT ANY CHANGE IN THE SET OF SENSITIVITY LEVELS ASSOCIATED WITH THE INFORMATION WHICH CAN BE TRANSMITTED OVER A MULTILEVEL COMMUNICATION CHANNEL OR COMPONENT. + Rationale COMMUNICATION CHANNELS AND COMPONENTS IN A NETWORK ARE ANALOGOUS TO COMMUNICATION CHANNELS AND I/O DEVICES IN STAND-ALONE SYSTEMS. THEY MUST BE DESIGNATED AS EITHER MUL- TILEVEL (I.E., ABLE TO DISTINGUISH AND MAINTAIN SEPARATION AMONG INFORMATION OF VARIOUS SENSITIVITY LEVELS) OR SINGLE- LEVEL. AS IN THE TCSEC, SINGLE-LEVEL DEVICES MAY ONLY BE ATTACHED TO SINGLE-LEVEL CHANNELS. THE LEVEL OR SET OF LEVELS OF INFORMATION THAT CAN BE SENT TO A COMPONENT OR OVER A COMMUNICATION CHANNEL SHALL ONLY CHANGE WITH THE KNOWLEDGE AND APPROVAL OF THE SECURITY OFFICERS (OR SYSTEM ADMINISTRATOR, IF THERE IS NO SECURITY OFFICER) OF THE NETWORK, AND OF THE AFFECTED COMPONENTS. THIS REQUIREMENT ENSURES THAT NO SIGNIFICANT SECURITY- RELEVANT CHANGES ARE MADE WITHOUT THE APPROVAL OF ALL AFFECTED PARTIES. 3.1.1.3.2.1 Exportation to Multilevel Devices + Statement from DoD 5200.28-STD WHEN THE TCB EXPORTS AN OBJECT TO A MULTILEVEL I/O DEVICE, THE SENSITIVITY LABEL ASSOCIATED WITH THAT OBJECT SHALL ALSO BE EXPORTED AND SHALL RESIDE ON THE SAME PHYSICAL MEDIUM AS THE EXPORTED INFORMATION AND SHALL BE IN THE SAME FORM (I.E., MACHINE-READABLE OR HUMAN-READABLE FORM). WHEN THE TCB EXPORTS OR IMPORTS AN OBJECT OVER A MULTILEVEL COMMUNI- CATIONS CHANNEL, THE PROTOCOL USED ON THAT CHANNEL SHALL PROVIDE FOR THE UNAMBIGUOUS PAIRING BETWEEN THE SENSITIVITY LABELS AND THE ASSOCIATED INFORMATION THAT IS SENT OR RECEIVED. + Interpretation THE COMPONENTS, INCLUDING HOSTS, OF A NETWORK SHALL BE INTERCONNECTED OVER "MULTILEVEL COMMUNICATION CHANNELS," MULTIPLE SINGLE-LEVEL COMMUNICATION CHANNELS, OR BOTH, WHEN- EVER THE INFORMATION IS TO BE PROTECTED AT MORE THAN A SIN- GLE SENSITIVITY LEVEL. THE PROTOCOL FOR ASSOCIATING THE SENSITIVITY LABEL AND THE EXPORTED INFORMATION SHALL PROVIDE THE ONLY INFORMATION NEEDED TO CORRECTLY ASSOCIATE A SENSI- TIVITY LEVEL WITH THE EXPORTED INFORMATION TRANSFERRED OVER THE MULTILEVEL CHANNEL BETWEEN THE NTCB PARTITIONS IN INDI- VIDUAL COMPONENTS. THIS PROTOCOL DEFINITION MUST SPECIFY THE REPRESENTATION AND SEMANTICS OF THE SENSITIVITY LABELS (I.E., THE MACHINE-READABLE LABEL MUST UNIQUELY REPRESENT THE SENSITIVITY LEVEL). THE "UNAMBIGUOUS" ASSOCIATION OF THE SENSITIVITY LEVEL WITH THE COMMUNICATED INFORMATION SHALL MEET THE SAME LEVEL OF ACCURACY AS THAT REQUIRED FOR ANY OTHER LABEL WITHIN THE NTCB, AS SPECIFIED IN THE CRITERION FOR LABEL INTEGRITY. THIS MAY BE PROVIDED BY PROTECTED AND HIGHLY RELIABLE DIRECT PHYSICAL LAYER CONNECTIONS, OR BY TRADITIONAL CRYPTOGRAPHIC LINK PROTECTION IN WHICH ANY ERRORS DURING TRANSMISSION CAN BE READILY DETECTED, OR BY USE OF A SEPARATE CHANNEL. + Rationale THIS PROTOCOL MUST SPECIFY THE REPRESENTATION AND SEMANTICS OF THE SENSITIVITY LABELS. SEE THE MANDATORY ACCESS CONTROL POLICIES SECTION IN APPENDIX B. THE MUL- TILEVEL DEVICE INTERFACE TO (UNTRUSTED) SUBJECTS MAY BE IMPLEMENTED EITHER BY THE INTERFACE OF THE REFERENCE MONI- TOR, PER SE, OR BY A MULTILEVEL SUBJECT (E.G., A "TRUSTED SUBJECT" AS DEFINED IN THE BELL-LAPADULA MODEL) THAT PRO- VIDES THE LABELS BASED ON THE INTERNAL LABELS OF THE NTCB PARTITION. THE CURRENT STATE OF THE ART LIMITS THE SUPPORT FOR MANDATORY POLICY THAT IS PRACTICAL FOR SECURE NETWORKS. REFERENCE MONITOR SUPPORT TO ENSURE THE CONTROL OVER ALL THE OPERATIONS OF EACH SUBJECT IN THE NETWORK MUST BE COMPLETELY PROVIDED WITHIN THE SINGLE NTCB PARTITION ON WHICH THAT SUB- JECT INTERFACES TO THE NTCB. THIS MEANS THAT THE ENTIRE PORTION OF THE "SECURE STATE" REPRESENTED IN THE SECURITY POLICY MODEL THAT MAY BE CHANGED BY TRANSITIONS INVOKED BY THIS SUBJECT MUST BE CONTAINED IN THE SAME COMPONENT. THE SECURE STATE OF AN NTCB PARTITION MAY BE AFFECTED BY EVENTS EXTERNAL TO THE COMPONENT IN WHICH THE NTCB PARTI- TION RESIDES (E.G., ARRIVAL OF A MESSAGE). THE EFFECT OCCURS ASYNCHRONUSLY AFTER BEING INITIATED BY AN EVENT IN ANOTHER COMPONENT OR PARTITION. FOR EXAMPLE, INDETERMINATE DELAYS MAY OCCUR BETWEEN THE INITIATION OF A MESSAGE IN ONE COMPONENT, THE ARRIVAL OF THE MESSAGE IN THE NTCB PARTITION IN ANOTHER COMPONENT, AND THE CORRESPONDING CHANGE TO THE SECURE STATE OF THE SECOND COMPONENT. SINCE EACH COMPONENT IS EXECUTING CONCURRENTLY, TO DO OTHERWISE WOULD REQUIRE SOME SORT OF NETWORK-WIDE CONTROL TO SYNCHRONIZE STATE TRAN- SITIONS, SUCH AS A GLOBAL NETWORK-WIDE CLOCK FOR ALL PROCES- SORS; IN GENERAL, SUCH DESIGNS ARE NOT PRACTICAL AND PROB- ABLY NOT EVEN DESIRABLE. THEREFORE, THE INTERACTION BETWEEN NTCB PARTITIONS IS RESTRICTED TO JUST COMMUNICATIONS BETWEEN PAIRS (AT LEAST LOGICALLY) OF DEVICES-MULTILEVEL DEVICES IF THE DEVICE(S) CAN SEND/RECEIVE DATA OF MORE THAN A SINGLE LEVEL. FOR BROADCAST CHANNELS THE PAIRS ARE THE SENDER AND INTENDED RECEIVER(S). HOWEVER, IF THE BROADCAST CHANNEL CARRIES MULTIPLE LEVELS OF INFORMATION, ADDITIONAL MECHANISM (E.G., CRYPTOCHECKSUM MAINTAINED BY THE TCB) MAY BE REQUIRED TO ENFORCE SEPARATION AND PROPER DELIVERY. A COMMON REPRESENTATION FOR SENSITIVITY LABELS IS NEEDED IN THE PROTOCOL USED ON THAT CHANNEL AND UNDERSTOOD BY BOTH THE SENDER AND RECEIVER WHEN TWO MULTILEVEL DEVICES (IN THIS CASE, IN TWO DIFFERENT COMPONENTS) ARE INTERCON- NECTED. EACH DISTINCT SENSITIVITY LEVEL OF THE OVERALL NET- WORK POLICY MUST BE REPRESENTED UNIQUELY IN THESE LABELS. WITHIN A MONOLITHIC TCB, THE ACCURACY OF THE SENSI- TIVITY LABELS IS GENERALLY ASSURED BY SIMPLE TECHNIQUES, E.G., VERY RELIABLE CONNECTIONS OVER VERY SHORT PHYSICAL CONNECTIONS, SUCH AS ON A SINGLE PRINTED CIRCUIT BOARD OR OVER AN INTERNAL BUS. IN MANY NETWORK ENVIRONMENTS THERE IS A MUCH HIGHER PROBABILITY OF ACCIDENTALLY OR MALICIOUSLY INTRODUCED ERRORS, AND THESE MUST BE PROTECTED AGAINST. 3.1.1.3.2.2 Exportation to Single-Level Devices + Statement from DoD 5200.28-STD SINGLE-LEVEL I/O DEVICES AND SINGLE-LEVEL COMMUNICATION CHANNELS ARE NOT REQUIRED TO MAINTAIN THE SENSITIVITY LABELS OF THE INFORMATION THEY PROCESS. HOWEVER, THE TCB SHALL INCLUDE A MECHANISM BY WHICH THE TCB AND AN AUTHORIZED USER RELIABLY COMMUNICATE TO DESIGNATE THE SINGLE SENSITIVITY LEVEL OF INFORMATION IMPORTED OR EXPORTED VIA SINGLE-LEVEL COMMUNICATION CHANNELS OR I/O DEVICES. + Interpretation WHENEVER ONE OR BOTH OF TWO DIRECTLY CONNECTED COM- PONENTS IS NOT TRUSTED TO MAINTAIN THE SEPARATION OF INFOR- MATION OF DIFFERENT SENSITIVITY LEVELS, OR WHENEVER THE TWO DIRECTLY CONNECTED COMPONENTS HAVE ONLY A SINGLE SENSITIVITY LEVEL IN COMMON, THE TWO COMPONENTS OF THE NETWORK SHALL COMMUNICATE OVER A SINGLE-LEVEL CHANNEL. SINGLE-LEVEL COM- PONENTS AND SINGLE-LEVEL COMMUNICATION CHANNELS ARE NOT REQUIRED TO MAINTAIN THE SENSITIVITY LABELS OF THE INFORMATION THEY PROCESS. HOWEVER, THE NTCB SHALL INCLUDE A RELIABLE COMMUNICATION MECHANISM BY WHICH THE NTCB AND AN AUTHORIZED USER OR A SUBJECT WITHIN AN NTCB PARTITION CAN DESIGNATE THE SINGLE SENSITIVITY LEVEL OF INFORMATION IMPORTED OR EXPORTED VIA SINGLE-LEVEL COMMUNICATION CHANNELS OR NETWORK COMPONENTS. + Rationale SINGLE-LEVEL COMMUNICATIONS CHANNELS AND SINGLE-LEVEL COMPONENTS IN NETWORKS ARE ANALOGOUS TO SINGLE LEVEL CHAN- NELS AND I/O DEVICES IN STAND-ALONE SYSTEMS IN THAT THEY ARE NOT TRUSTED TO MAINTAIN THE SEPARATION OF INFORMATION OF DIFFERENT SENSITIVITY LEVELS. THE LABELS ASSOCIATED WITH DATA TRANSMITTED OVER THOSE CHANNELS AND BY THOSE COMPONENTS ARE THEREFORE IMPLICIT; THE NTCB ASSOCIATES LABELS WITH THE DATA BECAUSE OF THE CHANNEL OR COMPONENT, NOT BECAUSE OF AN EXPLICIT PART OF THE BIT STREAM. NOTE THAT THE SENSITIVITY LEVEL OF ENCRYPTED INFORMATION IS THE LEVEL OF THE CIPHER- TEXT RATHER THAN THE ORIGINAL LEVEL(S) OF THE PLAINTEXT. 3.1.1.3.2.3 Labeling Human-Readable Output + Statement from DoD 5200.28-STD THE ADP SYSTEM ADMINISTRATOR SHALL BE ABLE TO SPECIFY THE PRINTABLE LABEL NAMES ASSOCIATED WITH EXPORTED SENSITIVITY LABELS. THE TCB SHALL MARK THE BEGINNING AND END OF ALL HUMAN-READABLE, PAGED, HARDCOPY OUTPUT (E.G., LINE PRINTER OUTPUT) WITH HUMAN-READABLE SENSITIVITY LABELS THAT PROP- ERLY1 REPRESENT THE SENSITIVITY OF THE OUTPUT. THE TCB SHALL, BY DEFAULT, MARK THE TOP AND BOTTOM OF EACH PAGE OF HUMAN-READABLE, PAGED, HARDCOPY OUTPUT (E.G., LINE PRINTER OUTPUT) WITH HUMAN-READABLE SENSITIVITY LABELS THAT PROP- ERLY1 REPRESENT THE SENSITIVITY OF THE PAGE. THE TCB SHALL, BY DEFAULT AND IN AN APPROPRIATE MANNER, MARK OTHER FORMS OF HUMAN READABLE OUTPUT (E.G., MAPS, GRAPHICS) WITH HUMAN- READABLE SENSITIVITY LABELS THAT PROPERLY1 REPRESENT THE SENSITIVITY OF THE OUTPUT. ANY OVERRIDE OF THESE MARKINGS DEFAULTS SHALL BE AUDITABLE BY THE TCB. + Interpretation THIS CRITERION IMPOSES NO REQUIREMENT TO A COMPONENT THAT PRODUCES NO HUMAN-READABLE OUTPUT. FOR THOSE THAT DO PRODUCE HUMAN-READABLE OUTPUT, EACH SENSITIVITY LEVEL THAT _________________________ 1 THE HIERARCHICAL CLASSIFICATION COMPONENT IN HUMAN- READABLE SENSITIVITY LABELS SHALL BE EQUAL TO THE GREATEST HIERARCHICAL CLASSIFICATION OF ANY OF THE IN- FORMATION IN THE OUTPUT THAT THE LABELS REFER TO; THE NON-HIERARCHICAL CATEGORY COMPONENT SHALL INCLUDE ALL OF THE NON-HIERARCHICAL CATEGORIES OF THE INFORMATION IN THE OUTPUT THE LABELS REFER TO, BUT NO OTHER NON- HIERARCHICAL CATEGORIES. IS DEFINED TO THE NETWORK SHALL HAVE A UNIFORM MEANING ACROSS ALL COMPONENTS. THE NETWORK ADMINISTRATOR, IN CON- JUNCTION WITH ANY AFFECTED COMPONENT ADMINISTRATOR, SHALL BE ABLE TO SPECIFY THE HUMAN-READABLE LABEL THAT IS ASSOCIATED WITH EACH DEFINED SENSITIVITY LEVEL. + Rationale THE INTERPRETATION IS A STRAIGHTFORWARD EXTENSION OF THE REQUIREMENT INTO THE CONTEXT OF A NETWORK SYSTEM AND PARTITIONED NTCB AS DEFINED FOR THESE NETWORK INTERPRETA- TIONS. 3.1.1.4 Mandatory Access Control + Statement from DoD 5200.28-STD THE TCB SHALL ENFORCE A MANDATORY ACCESS CONTROL POLICY OVER ALL SUBJECTS AND STORAGE OBJECTS UNDER ITS CONTROL (E.G., PROCESSES, FILES, SEGMENTS, DEVICES). THESE SUBJECTS AND OBJECTS SHALL BE ASSIGNED SENSITIVITY LABELS THAT ARE A COM- BINATION OF HIERARCHICAL CLASSIFICATION LEVELS AND NON- HIERARCHICAL CATEGORIES, AND THE LABELS SHALL BE USED AS THE BASIS FOR MANDATORY ACCESS CONTROL DECISIONS. THE TCB SHALL BE ABLE TO SUPPORT TWO OR MORE SUCH SENSITIVITY LEVELS. (SEE THE MANDATORY ACCESS CONTROL INTERPRETATIONS.) THE FOLLOWING REQUIREMENTS SHALL HOLD FOR ALL ACCESSES BETWEEN SUBJECTS AND OBJECTS CONTROLLED BY THE TCB: A SUBJECT CAN READ AN OBJECT ONLY IF THE HIERARCHICAL CLASSIFICATION IN THE SUBJECT'S SENSITIVITY LEVEL IS GREATER THAN OR EQUAL TO THE HIERARCHICAL CLASSIFICATION OF THE OBJECT'S SENSITIVITY LEVEL AND THE NON-HIERARCHICAL CATEGORIES IN THE SUBJECT'S SENSITIVITY LEVEL INCLUDE ALL THE NON-HIERARCHICAL CATEGORIES IN THE OBJECT'S SENSITIVITY LEVEL. A SUBJECT CAN WRITE AN OBJECT ONLY IF THE HIERARCHICAL CLASSIFICATION IN THE SUBJECT'S SENSITIVITY LEVEL IS LESS THAN OR EQUAL TO THE HIERARCHICAL CLASSIFICATION OF THE OBJECT'S SENSITIVITY LEVEL AND THE NON-HIERARCHICAL CATEGORIES IN THE SUBJECT'S SENSITIVITY LEVEL ARE INCLUDED IN THE NON-HIERARCHICAL CATEGORIES IN THE OBJECT'S SENSITIVITY LEVEL. IDENTIFICATION AND AUTHENTICATION DATA SHALL BE USED BY THE TCB TO AUTHEN- TICATE THE USER'S IDENTITY AND TO ENSURE THAT THE SENSI- TIVITY LEVEL AND AUTHORIZATION OF SUBJECTS EXTERNAL TO THE TCB THAT MAY BE CREATED TO ACT ON BEHALF OF THE INDIVIDUAL USER ARE DOMINATED BY THE CLEARANCE AND AUTHORIZATION OF THAT USER. + Interpretation EACH PARTITION OF THE NTCB EXERCISES MANDATORY ACCESS CONTROL POLICY OVER ALL SUBJECTS AND OBJECTS IN ITS COM- PONENT THAT ARE UNDER ITS CONTROL. IN A NETWORK, THE RESPONSIBILITY OF AN NTCB PARTITION ENCOMPASSES ALL MANDA- TORY ACCESS CONTROL FUNCTIONS IN ITS COMPONENT THAT WOULD BE REQUIRED OF A TCB IN A STAND-ALONE SYSTEM. IN PARTICULAR, SUBJECTS AND OBJECTS USED FOR COMMUNICATION WITH OTHER COM- PONENTS ARE UNDER THE CONTROL OF THE NTCB PARTITION. MANDA- TORY ACCESS CONTROL INCLUDES SECRECY AND INTEGRITY CONTROL TO THE EXTENT THAT THE NETWORK SPONSOR HAS DESCRIBED IN THE OVERALL NETWORK SECURITY POLICY. CONCEPTUAL ENTITIES ASSOCIATED WITH COMMUNICATION BETWEEN TWO COMPONENTS, SUCH AS SESSIONS, CONNECTIONS AND VIRTUAL CIRCUITS, MAY BE THOUGHT OF AS HAVING TWO ENDS, ONE IN EACH COMPONENT, WHERE EACH END IS REPRESENTED BY A LOCAL OBJECT. COMMUNICATION IS VIEWED AS AN OPERATION THAT COPIES INFORMATION FROM AN OBJECT AT ONE END OF A COMMUNICATION PATH TO AN OBJECT AT THE OTHER END. TRANSIENT DATA-CARRYING ENTITIES, SUCH AS DATAGRAMS AND PACKETS, EXIST EITHER AS INFORMATION WITHIN OTHER OBJECTS, OR AS A PAIR OF OBJECTS, ONE AT EACH END OF THE COMMUNICATION PATH. THE REQUIREMENT FOR "TWO OR MORE" SENSITIVITY LEVELS CAN BE MET BY EITHER SECRECY OR INTEGRITY LEVELS. WHEN THERE IS A MANDATORY INTEGRITY POLICY, THE STATED REQUIRE- MENTS FOR READING AND WRITING ARE GENERALIZED TO: A SUBJECT CAN READ AN OBJECT ONLY IF THE SUBJECT'S SENSITIVITY LEVEL DOMINATES THE OBJECT'S SENSITIVITY LEVEL, AND A SUBJECT CAN WRITE AN OBJECT ONLY IF THE OBJECT'S SENSITIVITY LEVEL DOM- INATES THE SUBJECT'S SENSITIVITY LEVEL. BASED ON THE INTEGRITY POLICY, THE NETWORK SPONSOR SHALL DEFINE THE DOMI- NANCE RELATION FOR THE TOTAL LABEL, FOR EXAMPLE, BY COMBIN- ING SECRECY AND INTEGRITY LATTICES. - + Rationale AN NTCB PARTITION CAN MAINTAIN ACCESS CONTROL ONLY OVER SUBJECTS AND OBJECTS IN ITS COMPONENT. ACCESS BY A SUBJECT IN ONE COMPONENT TO INFORMATION CONTAINED IN AN OBJECT IN ANOTHER COMPONENT REQUIRES THE CREATION OF A SUBJECT IN THE REMOTE COMPONENT WHICH ACTS AS A SURROGATE FOR THE FIRST SUBJECT. THE MANDATORY ACCESS CONTROLS MUST BE ENFORCED AT THE INTERFACE OF THE REFERENCE MONITOR (VIZ. THE MECHANISM THAT CONTROLS PHYSICAL PROCESSING RESOURCES) FOR EACH NTCB PARTI- TION. THIS MECHANISM CREATES THE ABSTRACTION OF SUBJECTS AND OBJECTS WHICH IT CONTROLS. SOME OF THESE SUBJECTS OUT- SIDE THE REFERENCE MONITOR, PER SE, MAY BE DESIGNATED TO IMPLEMENT PART OF AN NTCB PARTITION'S MANDATORY POLICY, E.G., BY USING THE ``TRUSTED SUBJECTS" DEFINED IN THE BELL- _________________________ - See, for example, Grohn, M. J., A Model of a Pro- _ _____ __ _ ___ tected Data Management System, ESD-TR-76-289, I. P. ______ ____ __________ ______ Sharp Assoc. Ltd., June, 1976; and Denning, D .E., Lunt, T. F., Neumann, P. G., Schell, R. R., Heckman, M. and Shockley, W., Secure Distributed Data Views, Secu- ______ ___________ ____ _____ ____ rity Policy and Interpretation for a Class A1 Multilev- ____ ______ ___ ______________ ___ _ _____ __ ________ el Secure Relational Database System,SRI International, __ ______ __________ ________ ______ November 1986. LAPADULA MODEL. THE PRIOR REQUIREMENTS ON EXPORTATION OF LABELED INFOR- MATION TO AND FROM I/O DEVICES ENSURE THE CONSISTENCY BETWEEN THE SENSITIVITY LABELS OF OBJECTS CONNECTED BY A COMMUNICATION PATH. AS NOTED IN THE INTRODUCTION, THE NET- WORK ARCHITECTURE MUST RECOGNIZE THE LINKAGE BETWEEN THE OVERALL MANDATORY NETWORK SECURITY POLICY AND THE CONNECTION ORIENTED ABSTRACTION. FOR EXAMPLE, INDIVIDUAL DATA-CARRYING ENTITIES SUCH AS DATAGRAMS CAN HAVE INDIVIDUAL SENSITIVITY LABELS THAT SUBJECT THEM TO MANDATORY ACCESS CONTROL IN EACH COMPONENT. THE ABSTRACTION OF A SINGLE-LEVEL CONNECTION IS REALIZED AND ENFORCED IMPLICITLY BY AN ARCHITECTURE WHILE A CONNECTION IS REALIZED BY SINGLE-LEVEL SUBJECTS THAT NECES- SARILY EMPLOY ONLY DATAGRAMS OF THE SAME LEVEL. THE FUNDAMENTAL TRUSTED SYSTEMS TECHNOLOGY PERMITS THE DAC MECHANISM TO BE DISTRIBUTED, IN CONTRAST TO THE REQUIRE- MENTS FOR MANDATORY ACCESS CONTROL. FOR NETWORKS THIS SEPARATION OF MAC AND DAC MECHANISMS IS THE RULE RATHER THAN THE EXCEPTION. THE SET OF TOTAL SENSITIVITY LABELS USED TO REPRESENT ALL THE SENSITIVITY LEVELS FOR THE MANDATORY ACCESS CONTROL (COMBINED DATA SECRECY AND DATA INTEGRITY) POLICY ALWAYS FORMS A PARTIALLY ORDERED SET. WITHOUT LOSS OF GENERALITY, THIS SET OF LABELS CAN ALWAYS BE EXTENDED TO FORM A LATTICE, BY INCLUDING ALL THE COMBINATIONS OF NON-HIERARCHICAL CATEGORIES. AS FOR ANY LATTICE, A DOMINANCE RELATION IS ALWAYS DEFINED FOR THE TOTAL SENSITIVITY LABELS. FOR ADMIN- ISTRATIVE REASONS IT MAY BE HELPFUL TO HAVE A MAXIMUM LEVEL WHICH DOMINATES ALL OTHERS. 3.1.2 Accountability _ _ _ ______________ 3.1.2.1 Identification and Authentication + Statement from DoD 5200.28-STD The TCB shall require users to identify themselves to it before beginning to perform any other actions that the TCB is expected to mediate. Furthermore, the TCB shall MAINTAIN AUTHENTICATION DATA THAT INCLUDES INFORMATION FOR VERIFYING THE IDENTIFY OF INDIVIDUAL USERS (E.G., PASSWORDS) AS WELL AS INFORMATION FOR DETERMINING THE CLEARANCE AND AUTHORIZA- TIONS OF INDIVIDUAL USERS. THIS DATA SHALL BE USED BY THE TCB TO AUTHENTICATE THE USER'S IDENTITY AND TO ENSURE THAT THE SENSITIVITY LEVEL AND AUTHORIZATION OF SUBJECTS EXTERNAL TO THE TCB THAT MAY BE CREATED TO ACT ON BEHALF OF THE INDI- VIDUAL USER ARE DOMINATED BY THE CLEARANCE AND AUTHORIZATION OF THAT USER. The TCB shall protect authentication data so that it cannot be accessed by any unauthorized user. The TCB shall be able to enforce individual accountability by providing the capability to uniquely identify each indivi- dual ADP system user. The TCB shall also provide the capability of associating this identity with all auditable actions taken by that individual. + Interpretation The requirement for identification and authentication of users is the same for a network system as for an ADP sys- tem. The identification and authentication may be done by the component to which the user is directly connected or some other component, such as an identification and authen- tication server. Available techniques, such as those described in the Password Guideline=, are generally also applicable in the network context. However, in cases where the NTCB is expected to mediate actions of a host (or other network component) that is acting on behalf of a user or group of users, the NTCB may employ identification and authentication of the host (or other component) in lieu of identification and authentication of an individual user, so long as the component identifier implies a list of specific users uniquely associated with the identifier at the time of its use for authentication. This requirement does not apply to internal subjects. Authentication information, including the identity of a user (once authenticated) may be passed from one component to another without reauthentication, so long as the NTCB protects (e.g., by encryption) the information from unau- thorized disclosure and modification. This protection shall provide at least a similar level of assurance (or strength of mechanism) as pertains to the protection of the authenti- cation mechanism and authentication data. + Rationale The need for accountability is not changed in the con- text of a network system. The fact that the NTCB is parti- tioned over a set of components neither reduces the need nor imposes new requirements. That is, individual accountabil- ity is still the objective. Also, in the context of a net- work system at the (C2) level or higher "individual accoun- tability" can be satisfied by identification of a host (or other component) so long as the requirement for traceability to individual users or a set of specific individual users with active subjects is satisfied. There is allowed to be an uncertainty in traceability because of elapsed time between changes in the group membership and the enforcement in the access control mechanisms. In addition, there is no need in a distributed processing system like a network to reauthen- ticate a user at each point in the network where a projec- tion of a user (via the subject operating on behalf of the user) into another remote subject takes place. _________________________ = Department of Defense Password Management Guide- __________ __ _______ ________ __________ _____ line, CSC-STD-002-85 ____ The passing of identifiers and/or authentication infor- mation from one component to another is usually done in sup- port to the implementation of the discretionary access con- trol (DAC). This support relates directly to the DAC regarding access by a user to a storage object in a dif- ferent NTCB partition than the one where the user was authenticated. Employing a forwarded identification implies additional reliance on the source and components along the path. IF THE AUTHENTICATED IDENTIFICATION IS USED AS THE BASIS OF DETERMINING A SENSITIVITY LABEL FOR A SUBJECT, IT MUST SATISFY THE LABEL INTEGRITY CRITERION. AN AUTHENTICATED IDENTIFICATION MAY BE FORWARDED BETWEEN COMPONENTS AND EMPLOYED IN SOME COMPONENT TO IDEN- TIFY THE SENSITIVITY LEVEL ASSOCIATED WITH A SUBJECT CREATED TO ACT ON BEHALF OF THE USER SO IDENTIFIED. 3.1.2.2 Audit _ _ _ _ _____ + Statement from DoD 5200.28-STD The TCB shall be able to create, maintain, and protect from modification or unauthorized access or destruction an audit trail of accesses to the objects it protects. The audit data shall be protected by the TCB so that read access to it is limited to those who are authorized for audit data. The TCB shall be able to record the following types of events: use of identification and authentication mechanisms, intro- duction of objects into a user's address space (e.g., file open, program initiation), deletion of objects, actions taken by computer operators and system administrators and/or system security officers, and other security relevant events. THE TCB SHALL ALSO BE ABLE TO AUDIT ANY OVERRIDE OF HUMAN-READABLE OUTPUT MARKINGS. For each recorded event, the audit record shall identify: date and time of the event, user, type of event, and success or failure of the event. For identification/authentication events the origin of request (e.g., terminal ID) shall be included in the audit record. For events that introduce an object into a user's address space and for object deletion events the audit record shall include the name of the object AND THE OBJECT'S SENSITIVITY LEVEL. The ADP system administrator shall be able to selectively audit the actions of any one or more users based on individual identify AND/OR OBJECT SENSITIVITY LEVEL. + Interpretation This criterion applies as stated. The sponsor must select which events are auditable. If any such events are not distinguishable by the NTCB alone (for example those identified in Part II), the audit mechanism shall provide an interface, which an authorized subject can invoke with parameters sufficient to produce an audit record. These audit records shall be distinguishable from those provided by the NTCB. In the context of a network system, "other security relevant events" (depending on network system architecture and network security policy) might be as fol- lows: 1. Identification of each access event (e.g., estab- lishing a connection or a connectionless association between processes in two hosts of the network) and its principal parameters (e.g., host identifiers of the two hosts involved in the access event and user identifier or host identifier of the user or host that is requesting the access event) 2. Identification of the starting and ending times of each access event using local time or global syn- chronized time 3. Identification of security-relevant exceptional con- ditions (e.g., potential violation of data integrity, such as misrouted datagrams) detected during the transactions between two hosts 4. Utilization of cryptographic variables 5. Changing the configuration of the network (e.g., a component leaving the network and rejoining) In addition, identification information should be included in appropriate audit trail records, as necessary, to allow association of all related (e.g., involving the same network event) audit trail records (e.g., at different hosts) with each other. Furthermore, a component of the network system may provide the required audit capability (e.g., storage, retrieval, reduction, analysis) for other components that do not internally store audit data but transmit the audit data to some designated collection com- ponent. Provisions shall be made to control the loss of audit data due to unavailability of resources. In the context of a network system, the "user's address space" is extended, for object introduction and deletion events, to include address spaces being employed on behalf of a remote user (or host). However, the focus remains on users in contrast to internal subjects as discussed in the DAC criterion. In addition, audit information must be stored in machine-readable form. + Rationale For remote users, the network identifiers (e.g., inter- net address) can be used as identifiers of groups of indivi- dual users (e.g., "all users at Host A") to eliminate the maintenance that would be required if individual identifica- tion of remote users was employed. In this class (C2), how- ever, it must be possible to identify (immediately or at some later time) the individuals represented by a group identifier. In all other respects, the interpretation is a straightforward extension of the criterion into the context of a network system. 3.1.3 Assurance _ _ _ _________ 3.1.3.1 Operational Assurance 3.1.3.1.1 System Architecture + Statement from DoD 5200.28-STD The TCB shall maintain a domain for its own execution that protects it from external interference or tampering (e.g., by modification of its code or data structures). Resources controlled by the TCB may be a defined subset of the sub- jects and objects in the ADP system. THE TCB SHALL MAINTAIN PROCESS ISOLATION THROUGH THE PROVISION OF DISTINCT ADDRESS SPACES UNDER ITS CONTROL. The TCB shall isolate the resources to be protected so that they are subject to the access control and auditing requirements. + Interpretation The system architecture criterion must be met individu- ally by all NTCB partitions. Implementation of the require- ment that the NTCB maintain a domain for its own execution is achieved by having each NTCB partition maintain a domain for its own execution. SINCE EACH COMPONENT IS ITSELF A DIS- TINCT DOMAIN IN THE OVERALL NETWORK SYSTEM, THIS ALSO SATIS- FIES THE REQUIREMENT FOR PROCESS ISOLATION THROUGH DISTINCT ADDRESS SPACES IN THE SPECIAL CASE WHERE A COMPONENT HAS ONLY A SINGLE SUBJECT. The subset of network resources over which the NTCB has control are the union of the sets of resources over which the NTCB partitions have control. Code and data structures belonging to the NTCB, transferred among NTCB subjects (i.e., subjects outside the reference monitor but inside the NTCB) belonging to different NTCB partitions, must be pro- tected against external interference or tampering. For example, a cryptographic checksum or physical means may be employed to protect user authentication data exchanged between NTCB partitions. Each NTCB partition provides isolation of RESOURCES (WITHIN ITS COMPONENT) TO BE PROTECTED IN accord with the network system architecture and security policy SO THAT "SUPPORTING ELEMENTS" (E.G., DAC AND USER IDENTIFICATION) FOR THE SECURITY MECHANISMS OF THE NETWORK SYSTEM ARE STRENGTHENED COMPARED TO C2, FROM AN ASSURANCE POINT OF VIEW, THROUGH THE PROVISION OF DISTINCT ADDRESS SPACES UNDER CONTROL OF THE NTCB. AS DISCUSSED IN THE DISCRETIONARY ACCESS CONTROL SEC- TION, THE DAC MECHANISM OF A NTCB PARTITION MAY BE IMPLE- MENTED AT THE INTERFACE OF THE REFERENCE MONITOR OR MAY BE DISTRIBUTED IN SUBJECTS THAT ARE PART OF THE NTCB IN THE SAME OR DIFFERENT COMPONENT. WHEN DISTRIBUTED IN NTCB SUB- JECTS (I.E., WHEN OUTSIDE THE REFERENCE MONITOR), THE ASSURANCE REQUIREMENTS FOR THE DESIGN AND IMPLEMENTATION OF THE DAC SHALL BE THOSE OF CLASS C2 FOR ALL NETWORKS OF CLASS C2 OR ABOVE. + Rationale The requirement for the protection of communications between NTCB partitions is specifically directed to subjects that are part of the NTCB partitions. Any requirements for such protection for the subjects that are outside the NTCB partitions are addressed in response to the integrity requirements of the security policy. THE PROVISION OF DISTINCT ADDRESS SPACES UNDER THE CON- TROL OF THE NTCB PROVIDES THE ABILITY TO SEPARATE SUBJECTS ACCORDING TO SENSITIVITY LEVEL. THIS REQUIREMENT IS INTRO- DUCED AT B1 SINCE IT IS AN ABSOLUTE NECESSITY IN ORDER TO IMPLEMENT MANDATORY ACCESS CONTROLS. 3.1.3.1.2 System Integrity + Statement from DoD 5200.28-STD Hardware and/or software features shall be provided that can be used to periodically validate the correct operation of the on-site hardware and firmware elements of the TCB. + Interpretation Implementation of the requirement is partly achieved by having hardware and/or software features that can be used to periodically validate the correct operation of the hardware and firmware elements of each component's NTCB partition. Features shall also be provided to validate the identity and correct operation of a component prior to its incorporation in the network system and throughout system operation. For example, a protocol could be designed that enables the com- ponents of the partitioned NTCB to exchange messages period- ically and validate each other's correct response. The pro- tocol shall be able to determine the remote entity's ability to respond. NTCB partitions shall provide the capability to report to network administrative personnel the failures detected in other NTCB partitions. Intercomponent protocols implemented within a NTCB shall be designed in such a way as to provide correct opera- tion in the case of failures of network communications or individual components. The allocation of MANDATORY AND dis- cretionary access control policy in a network may require communication between trusted subjects that are part of the NTCB partitions in different components. This communication is normally implemented with a protocol between the subjects as peer entities. Incorrect access within a component shall not result from failure of an NTCB partition to communicate with other components. + Rationale The first paragraph of the interpretation is a straightforward extension of the requirement into the con- text of a network system and partitioned NTCB as defined for these network criteria. NTCB protocols should be robust enough so that they permit the system to operate correctly in the case of local- ized failure. The purpose of this protection is to preserve the integrity of the NTCB itself. It is not unusual for one or more components in a network to be inoperative at any time, so it is important to minimize the effects of such failures on the rest of the network. Additional integrity and denial of service issues are addressed in Part II. 3.1.3.2 Life-Cycle Assurance 3.1.3.2.1 Security Testing + Statement from DoD 5200.28-STD The security mechanisms of the ADP system shall be tested and found to work as claimed in the system documentation. A TEAM OF INDIVIDUALS WHO THOROUGHLY UNDERSTAND THE SPECIFIC IMPLEMENTATION OF THE TCB SHALL SUBJECT ITS DESIGN DOCUMEN- TATION, SOURCE CODE, AND OBJECT CODE TO THROUGH ANALYSIS AND TESTING. THEIR OBJECTIVES SHALL BE: TO UNCOVER ALL DESIGN AND IMPLEMENTATION FLAWS THAT WOULD PERMIT A SUBJECT EXTER- NAL TO THE TCB TO READ, CHANGE, OR DELETE DATA NORMALLY DENIED UNDER THE MANDATORY OR DISCRETIONARY SECURITY POLICY ENFORCED BY THE TCB; AS WELL AS TO ASSURE THAT NO SUBJECT (WITHOUT AUTHORIZATION TO DO SO) IS ABLE TO CAUSE THE TCB TO ENTER A STATE SUCH THAT IT IS UNABLE TO RESPOND TO COMMUNI- CATIONS INITIATED BY OTHER USERS. ALL DISCOVERED FLAWS SHALL BE REMOVED OR NEUTRALIZED AND THE TCB RETESTED TO DEMON- STRATE THAT THEY HAVE BEEN ELIMINATED AND THAT NEW FLAWS HAVE NOT BEEN INTRODUCED. (See the Security Testing Guide- lines.) + Interpretation Testing of a component will require a testbed that exercises the interfaces and protocols of the component including tests under exceptional conditions. The testing of a security mechanism of the network system for meeting this criterion shall be an integrated testing procedure involving all components containing an NTCB partition that implement the given mechanism. This integrated testing is additional to any individual component tests involved in the evaluation of the network system. The sponsor should iden- tify the allowable set of configurations including the sizes of the networks. Analysis or testing procedures and tools shall be available to test the limits of these configura- tions. A change in configuration within the allowable set of configurations does not require retesting. THE TESTING OF EACH COMPONENT WILL INCLUDE THE INTRO- DUCTION OF SUBJECTS EXTERNAL TO THE NTCB PARTITION FOR THE COMPONENT THAT WILL ATTEMPT TO READ, CHANGE, OR DELETE DATA NORMALLY DENIED. IF THE NORMAL INTERFACE TO THE COMPONENT DOES NOT PROVIDE A MEANS TO CREATE THE SUBJECTS NEEDED TO CONDUCT SUCH A TEST, THEN THIS PORTION OF THE TESTING SHALL USE A SPECIAL VERSION OF THE UNTRUSTED SOFTWARE FOR THE COM- PONENT THAT RESULTS IN SUBJECTS THAT MAKE SUCH ATTEMPTS. THE RESULTS SHALL BE SAVED FOR TEST ANALYSIS. SUCH SPECIAL VERSIONS SHALL HAVE AN NTCB PARTITION THAT IS IDENTICAL TO THAT FOR THE NORMAL CONFIGURATION OF THE COMPONENT UNDER EVALUATION. THE TESTING OF THE MANDATORY CONTROLS SHALL INCLUDE TESTS TO DEMONSTRATE THAT THE LABELS FOR INFORMATION IMPORTED AND/OR EXPORTED TO/FROM THE COMPONENT ACCURATELY REPRESENT THE LABELS MAINTAINED BY THE NTCB PARTITION FOR THE COMPONENT FOR USE AS THE BASIS FOR ITS MANDATORY ACCESS CONTROL DECISIONS. THE TESTS SHALL INCLUDE EACH TYPE OF DEVICE, WHETHER SINGLE-LEVEL OR MULTILEVEL, SUPPORTED BY THE COMPONENT. + Rationale THE PHRASE "NO SUBJECT (WITHOUT AUTHORIZATION TO DO SO) IS ABLE TO CAUSE THE TCB TO ENTER A STATE SUCH THAT IT IS UNABLE TO RESPOND TO COMMUNICATIONS INITIATED BY OTHER USERS" RELATES TO THE SECURITY SERVICES (PART II OF THIS TNI) FOR THE DENIAL OF SERVICE PROBLEM, AND TO CORRECTNESS OF THE PROTOCOL IMPLEMENTATIONS. Testing is AN IMPORTANT method available in this evaluation division to gain any assurance that the security mechanisms perform their intended function. A MAJOR PURPOSE OF TESTING IS TO DEMONSTRATE THE SYSTEM'S RESPONSE TO INPUTS TO THE NTCB PARTITION FROM UNTRUSTED (AND POSSIBLY MALI- CIOUS) SUBJECTS. IN CONTRAST TO GENERAL PURPOSE SYSTEMS THAT ALLOW FOR THE DYNAMIC CREATION OF NEW PROGRAMS AND THE INTRODUCTIONS OF NEW PROCESSES (AND HENCE NEW SUBJECTS) WITH USER SPECI- FIED SECURITY PROPERITIES, MANY NETWORK COMPONENTS HAVE NO METHOD FOR INTRODUCING NEW PROGRAMS AND/OR PROCESSES DURING THEIR NORMAL OPERATION. THEREFORE, THE PROGRAMS NECESSARY FOR THE TESTING MUST BE INTRODUCED AS SPECIAL VERSIONS OF THE SOFTWARE RATHER THAN AS THE RESULT OF NORMAL INPUTS BY THE TEST TEAM. HOWEVER, IT MUST BE INSURED THAT THE NTCB PARTITION USED FOR SUCH TESTS IS IDENTICAL TO THE ONE UNDER EVALUATION. SENSITIVITY LABELS SERVE A CRITICAL ROLE IN MAINTAINING THE SECURITY OF THE MANDATORY ACCESS CONTROLS IN THE NET- WORK. ESPECIALLY IMPORTANT TO NETWORK SECURITY IS THE ROLE OF THE LABELS FOR INFORMATION COMMUNICATED BETWEEN COM- PONENTS - EXPLICIT LABELS FOR MULTILEVEL DEVICES AND IMPLI- CIT LABELS FOR SINGLE-LEVEL DEVICES. THEREFORE THE TESTING FOR CORRECT LABELS IS HIGHLIGHTED. 3.1.3.2.2 Design Specification and Verification + Statement from DoD 5200.28-STD AN INFORMAL OR FORMAL MODEL OF THE SECURITY POLICY SUPPORTED BY THE TCB SHALL BE MAINTAINED OVER THE LIFE CYCLE OF THE ADP SYSTEM AND DEMONSTRATED TO BE CONSISTENT WITH ITS AXIOMS. + Interpretation THE OVERALL NETWORK SECURITY POLICY EXPRESSED IN THIS MODEL WILL PROVIDE THE BASIS FOR THE MANDATORY ACCESS CON- TROL POLICY EXERCISED BY THE NTCB OVER SUBJECTS AND STORAGE OBJECTS IN THE ENTIRE NETWORK. THE POLICY WILL ALSO BE THE BASIS FOR THE DISCRETIONARY ACCESS CONTROL POLICY EXERCISED BY THE NTCB TO CONTROL ACCESS OF NAMED USERS TO NAMED OBJECTS. DATA INTEGRITY REQUIREMENTS ADDRESSING THE EFFECTS OF UNAUTHORIZED MSM NEED NOT BE INCLUDED IN THIS MODEL. THE OVERALL NETWORK POLICY MUST BE DECOMPOSED INTO POLICY ELE- MENTS THAT ARE ALLOCATED TO APPROPRIATE COMPONENTS AND USED AS THE BASIS FOR THE SECURITY POLICY MODEL FOR THOSE COM- PONENTS. THE LEVEL OF ABSTRACTION OF THE MODEL, AND THE SET OF SUBJECTS AND OBJECTS THAT ARE EXPLICITLY REPRESENTED IN THE MODEL, WILL BE AFFECTED BY THE NTCB PARTITIONING. SUBJECTS AND OBJECTS MUST BE REPRESENTED EXPLICITLY IN THE MODEL FOR THE PARTITION IF THERE IS SOME NETWORK COMPONENT WHOSE NTCB PARTITION EXERCISES ACCESS CONTROL OVER THEM. THE MODEL SHALL BE STRUCTURED SO THAT THE AXIOMS AND ENTITIES APPLICA- BLE TO INDIVIDUAL NETWORK COMPONENTS ARE MANIFEST. GLOBAL NETWORK POLICY ELEMENTS THAT ARE ALLOCATED TO COMPONENTS SHALL BE REPRESENTED BY THE MODEL FOR THAT COMPONENT. + Rationale THE TREATMENT OF THE MODEL DEPENDS TO A GREAT EXTENT ON THE DEGREE OF INTEGRATION OF THE COMMUNICATIONS SERVICE INTO A DISTRIBUTED SYSTEM. IN A CLOSELY COUPLED DISTRIBUTED SYS- TEM, ONE MIGHT USE A MODEL THAT CLOSELY RESEMBLES ONE APPROPRIATE FOR A STAND-ALONE COMPUTER SYSTEM. IN OTHER CASES, THE MODEL OF EACH PARTITION WILL BE EXPECTED TO SHOW THE ROLE OF THE NTCB PARTITION IN EACH KIND OF COMPONENT. IT WILL MOST LIKELY CLARIFY THE MODEL, ALTHOUGH NOT PART OF THE MODEL, TO SHOW ACCESS RESTRICTIONS IMPLIED BY THE SYSTEM DESIGN; FOR EXAMPLE, SUBJECTS REPRESENTING PROTOCOL ENTITIES MIGHT HAVE ACCESS ONLY TO OBJECTS CONTAINING DATA UNITS AT THE SAME LAYER OF PROTOCOL. THE ALLOCATION OF SUBJECTS AND OBJECTS TO DIFFERENT PROTO- COL LAYERS IS A PROTOCOL DESIGN CHOICE WHICH NEED NOT BE REFLECTED IN THE SECURITY POLICY MODEL. 3.1.4 Documentation. _ _ _ _____________ 3.1.4.1 Security Features User's Guide + Statement from DoD 5200.28-STD A single summary, chapter, or manual in user documentation shall describe the protection mechanisms provided by the TCB, interpretations on their use, and how they interact with one another. + Interpretation This user documentation describes user visible protec- tion mechanisms at the global (network system) level and at the user interface of each component, and the interaction among these. + Rationale The interpretation is an extension of the requirement into the context of a network system as defined for these network criteria. Documentation of protection mechanisms provided by individual components is required by the cri- teria for trusted computer systems that are applied as appropriate for the individual components. 3.1.4.2 Trusted Facility Manual + Statement from DoD 5200.28-STD A manual addressed to the ADP system administrator shall present cautions about functions and privileges that should be controlled when running a secure facility. The procedures for examining and maintaining the audit files as well as the detailed audit record structure for each type of audit event shall be given. THE MANUAL SHALL DESCRIBE THE OPERATOR AND ADMINISTRATOR FUNCTIONS RELATED TO SECURITY, TO INCLUDE CHANGING THE SECURITY CHARACTERISTICS OF A USER. IT SHALL PROVIDE INTERPRETATIONS ON THE CONSISTENT AND EFFECTIVE USE OF THE PROTECTION FEATURES OF THE SYSTEM, HOW THEY INTERACT, HOW TO SECURELY GENERATE A NEW TCB, AND FACILITY PROCEDURES, WARNINGS, AND PRIVILEGES THAT NEED TO BE CONTROLLED IN ORDER TO OPERATE THE FACILITY IN A SECURE MANNER. + Interpretation This manual shall contain specifications and procedures to assist the system administrator(s) maintain cognizance of the network configuration. These specifications and pro- cedures shall address the following: 1. The hardware configuration of the network itself; 2. The implications of attaching new components to the network; 3. The case where certain components may periodically leave the network (e.g., by crashing, or by being disconnected) and then rejoin; 4. Network configuration aspects that can impact the security of the network system; (For example, the manual should describe for the network system administrator the interconnections among components that are consistent with the overall network system architecture.) 5. Loading or modifying NTCB software or firmware (e.g., down-line loading). The physical and administrative environmental controls shall be specified. Any assumptions about security of a given network should be clearly stated (e.g., the fact that all communications links must be physically protected to a certain level). + Rationale There may be multiple system administrators with diverse responsibilities. The technical security measures described by these criteria must be used in conjunction with other forms of security in order to achieve security of the network. Additional forms include administrative security, physical security, emanations security, etc. Extension of this criterion to cover configuration aspects of the network is needed because, for example, proper interconnection of components is typically essential to achieve a correct realization of the network architec- ture. AS MENTIONED IN THE SECTION ON LABEL INTEGRITY, cryp- tography is one common mechanism employed to protect commun- ication circuits. Encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity of the ciphertext is generally lower than the cleartext. If encryption methodologies are employed, they shall be approved by the National Security Agency (NSA). The encryption algorithm and its implementation are outside the scope of these interpretations. This algorithm and implementation may be implemented in a separate device or may be a function of a subject in a component not dedi- cated to encryption. Without prejudice, either implementa- tion packaging is referred to as an encryption mechanism herein. 3.1.4.3 Test Documentation + Statement from DoD 5200.28-STD The system developer shall provide to the evaluators a docu- ment that describes the test plan, test procedures that show how the security mechanisms were tested, and results of the security mechanisms' functional testing. + Interpretation The "system developer" is interpreted as "the network system sponsor". The description of the test plan should establish the context in which the testing was or should be conducted. The description should identify any additional test components that are not part of the system being evaluated. This includes a description of the test-relevant functions of such test components and a description of the interfacing of those test components to the system being evaluated. The description of the test plan should also demonstrate that the tests adequately cover the network security policy. The tests should include the features described in the System Architecture and the System Integrity sections. The tests should also include network configuration and sizing. + Rationale The entity being evaluated may be a networking subsys- tem (see Appendix A) to which other components must be added to make a complete network system. In that case, this interpretation is extended to include contextual definition because, at evaluation time, it is not possible to validate the test plans without the description of the context for testing the networking subsystem. 3.1.4.4 Design Documentation + Statement from DoD 5200.28-STD Documentation shall be available that provides a description of the manufacturer's philosophy of protection and an expla- nation of how this philosophy is translated into the TCB. If the TCB is composed of distinct modules, the interfaces between these modules shall be described. AN INFORMAL OR FORMAL DESCRIPTION OF THE SECURITY POLICY MODEL ENFORCED BY THE TCB SHALL BE AVAILABLE AND AN EXPLANATION PROVIDED TO SHOW THAT IT IS SUFFICIENT TO ENFORCE THE SECURITY POLICY. THE SPECIFIC TCB PROTECTION MECHANISMS SHALL BE IDENTIFIED AND AN EXPLANATION GIVEN TO SHOW THAT THEY SATISFY THE MODEL. + Interpretation Explanation of how the sponsor's philosophy of protec- tion is translated into the NTCB shall include a description of how the NTCB is partitioned. The security policy also shall be stated. The description of the interfaces between the NTCB modules shall include the interface(s) between NTCB partitions and modules within the partitions if the modules exist. The sponsor shall describe the security architecture and design, including the allocation of security require- ments among components. Appendix A addresses component evaluation issues. AS STATED IN THE INTRODUCTION TO DIVISION B, THE SPON- SOR MUST DEMONSTRATE THAT THE NTCB EMPLOYS THE REFERENCE MONITOR CONCEPT. THE SECURITY POLICY MODEL MUST BE A MODEL FOR A REFERENCE MONITOR. THE SECURITY POLICY MODEL FOR EACH PARTITION IMPLEMENT- ING A REFERENCE MONITOR SHALL FULLY REPRESENT THE ACCESS CONTROL POLICY SUPPORTED BY THE PARTITION, INCLUDING THE DISCRETIONARY AND MANDATORY SECURITY POLICY FOR SECRECY AND/OR INTEGRITY. FOR THE MANDATORY POLICY THE SINGLE DOMI- NANCE RELATION FOR SENSITIVITY LABELS, INCLUDING SECRECY AND/OR INTEGRITY COMPONENTS, SHALL BE PRECISELY DEFINED. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system as defined for this network interpretation. Other documenta- tion, such as description of components and description of operating environment(s) in which the networking subsystem or network system is designed to function, is required else- where, e.g., in the Trusted Facility Manual. In order to be evaluated, a network must possess a coherent Network Security Architecture and Design. (Inter- connection of components that do not adhere to such a single coherent Network Security Architecture is addressed in the Interconnection of Accredited AIS, Appendix C.) The Network Security Architecture must address the security-relevant policies, objectives, and protocols. The Network Security Design specifies the interfaces and services that must be incorporated into the network so that it can be evaluated as a trusted entity. There may be multiple designs that con- form to the same architecture but are more or less incompa- tible and non-interoperable (except through the Interconnec- tion Rules). Security related mechanisms requiring coopera- tion among components are specified in the design in terms of their visible interfaces; mechanisms having no visible interfaces are not specified in this document but are left as implementation decisions. The Network Security Architecture and Design must be available from the network sponsor before evaluation of the network, or any component, can be undertaken. The Network Security Architecture and Design must be sufficiently com- plete, unambiguous, and free from obvious flaws to permit the construction or assembly of a trusted network based on the structure it specifies. When a component is being designed or presented for evaluation, or when a network assembled from components is assembled or presented for evaluation, there must be a priori evidence that the Network security Architecture and Design are satisfied. That is, the components can be assem- bled into a network that conforms in every way with the Net- work Security Architecture and Design to produce a physical realization that is trusted to the extent that its evalua- tion indicates. In order for a trusted network to be constructed from components that can be built independently, the Network Security Architecture and Design must completely and unambi- guously define the security functionality of components as well as the interfaces between or among components. The Network Security Architecture and Design must be evaluated to determine that a network constructed to its specifica- tions will in fact be trusted, that is, it will be evaluat- able under these interpretations. THE TERM "MODEL" IS USED IN SEVERAL DIFFERENT WAYS IN A NETWORK CONTEXT, E.G., A "PROTOCOL REFERENCE MODEL," A "FOR- MAL NETWORK MODEL," ETC. ONLY THE "SECURITY POLICY MODEL" IS ADDRESSED BY THIS REQUIREMENT AND IS SPECIFICALLY INTENDED TO MODEL THE INTERFACE, VIZ., "SECURITY PERIMETER," OF THE REFERENCE MONITOR AND MUST MEET ALL THE REQUIREMENTS DEFINED IN THE TCSEC. IT MUST BE SHOWN THAT ALL PARTS OF THE TCB ARE A VALID INTERPRETATION OF THE SECURITY POLICY MODEL, I.E., THAT THERE IS NO CHANGE TO THE SECURE STATE EXCEPT AS REPRESENTED BY THE MODEL. 3.2 CLASS (B2): STRUCTURED PROTECTION _ _ _____ __ __________ __________ IN CLASS (B2) NETWORK SYSTEMS, THE NTCB IS BASED ON A CLEARLY DEFINED AND DOCUMENTED FORMAL SECU- RITY POLICY MODEL THAT REQUIRES THE DISCRETIONARY AND MANDATORY ACCESS CONTROL ENFORCEMENT FOUND IN CLASS (B1) NETWORK SYSTEMS TO BE EXTENDED TO ALL SUBJECTS AND OBJECTS IN THE NETWORK SYSTEM. IN ADDITION, COVERT CHANNELS ARE ADDRESSED. THE NTCB MUST BE CAREFULLY STRUCTURED INTO PROTECTION- CRITICAL AND NON-PROTECTION-CRITICAL ELEMENTS. THE NTCB INTERFACE IS WELL-DEFINED, AND THE NTCB DESIGN AND IMPLEMENTATION ENABLE IT TO BE SUB- JECTED TO MORE THOROUGH TESTING AND MORE COMPLETE REVIEW. AUTHENTICATION MECHANISMS ARE STRENGTHENED, TRUSTED FACILITY MANAGEMENT IS PRO- VIDED IN THE FORM OF SUPPORT FOR SYSTEM ADMINIS- TRATOR AND OPERATOR FUNCTIONS, AND STRINGENT CON- FIGURATION MANAGEMENT CONTROLS ARE IMPOSED. THE SYSTEM IS RELATIVELY RESISTANT TO PENETRATION. THE FOLLOWING ARE MINIMAL REQUIREMENTS FOR SYSTEM ASSIGNED A CLASS (B2) RATING. 3.2.1 Security Policy _ _ _ ________ ______ + Statement from DoD 5200.28-STD Implied from the Introduction to the TCSEC. + Interpretation The network sponsor shall describe the overall network security policy enforced by the NTCB. At a minimum, this policy shall include the discretionary and mandatory requirements applicable to this class. The policy may require data secrecy, or data integrity, or both. The pol- icy is an access control policy having two primary com- ponents: mandatory and discretionary. The policy shall include a discretionary policy for protecting the informa- tion being processed based on the authorizations of indivi- duals, users, or groups of users. This access control pol- icy statement shall describe the requirements on the network to prevent or detect "reading or destroying" sensitive information by unauthorized users or errors. The mandatory policy must define the set of distinct sensitivity levels that it supports. For the Class B1 or above the mandatory policy shall be based on the labels associated with the information that reflects its sensitivity with respect to secrecy and/or integrity, where applicable, and labels asso- ciated with users to reflect their authorization to access such information. Unauthorized users include both those that are not authorized to use the network at all (e.g., a user attempting to use a passive or active wire tap) or a legitimate user of the network who is not authorized to access a specific piece of information being protected. Note that "users" does not include "operators," "system programmers," "technical control officers," "system security officers," and other system support personnel. They are distinct from users and are subject to the Trusted Facility Manual and the System Architecture requirements. Such indi- viduals may change the system parameters of the network sys- tem, for example, by defining membership of a group. These individuals may also have the separate role of users. SECRECY POLICY: The network sponsor shall define the form of the discretionary and mandatory secrecy policy that is enforced in the network to prevent unauthorized users from reading the sensitive infor- mation entrusted to the network. DATA INTEGRITY POLICY: The network sponsor shall define the discretionary and mandatory integrity policy to prevent unauthorized users from modifying, viz., writing, sensitive information. The defini- tion of data integrity presented by the network sponsor refers to the requirement that the informa- tion has not been subjected to unauthorized modifi- cation in the network. The mandatory integrity pol- icy enforced by the NTCB cannot, in general, prevent modification while information is being transmitted between components. However, an integrity sensi- tivity label may reflect the confidence that the information has not been subjected to transmission errors because of the protection afforded during transmission. This requirement is distinct from the requirement for label integrity. + Rationale The word "sponsor" is used in place of alternatives (such as "vendor," "architect," "manufacturer," and "developer") because the alternatives indicate people who may not be available, involved, or relevant at the time that a network system is proposed for evaluation. A trusted network is able to control both the reading and writing of shared sensitive information. Control of writing is used to protect against destruction of informa- tion. A network normally is expected to have policy require- ments to protect both the secrecy and integrity of the information entrusted to it. In a network the integrity is frequently as important or more important than the secrecy requirements. Therefore the secrecy and/or integrity policy to be enforced by the network must be stated for each net- work regardless of its evaluation class. The assurance that the policy is faithfully enforced is reflected in the evaluation class of the network. This control over modification is typically used to protect information so that it may be relied upon and to control the potential harm that would result if the informa- tion were corrupted. The overall network policy require- ments for integrity includes the protection for data both while being processed in a component and while being transmitted in the network. The access control policy enforced by the NTCB relates to the access of subjects to objects within each component. Communications integrity addressed within Part II relates to information while being transmitted. The mandatory integrity policy (at class B1 and above) in some architectures may be useful in supporting the link- age between the connection oriented abstraction introduced in the Introduction and the individual components of the network. For example, in a key distribution center for end-to-end encryption, a distinct integrity category may be assigned to isolate the key generation code and data from possible modification by other supporting processes in the same component, such as operator interfaces and audit. The mandatory integrity policy for some architecture may define an integrity sensitivity label that reflects the specific requirements for ensuring that information has not been subject to random errors in excess of a stated limit nor to unauthorized message stream modification (MSM) -. The specific metric associated with an integrity sensitivity label will generally reflect the intended applications of the network. 3.2.1.1 Discretionary Access Control + Statement from DoD 5200.28-STD The TCB shall define and control access between named users and named objects (e.g., files and programs) in the ADP sys- tem. The enforcement mechanism (e.g., self/group/public _________________________ - See Voydock, Victor L. and Stephen T. Kent, "Secu- rity Mechanisms in High-Level Network Protocols," Com- ___ puting Surveys, Vol. 15, No. 2, June 1983, pp 135-171. ______ _______ controls, access control lists) shall allow users to specify and control sharing of those objects by named individuals or defined groups of individuals, or both, and shall provide controls to limit propagation of access rights. The discre- tionary access control mechanism shall, either by explicit user action or by default, provide that objects are pro- tected from unauthorized access. These access controls shall be capable of including or excluding access to the granularity of a single user. Access permission to an object by users not already possessing access permission shall only be assigned by authorized users. + Interpretation The discretionary access control (DAC) mechanism(s) may be distributed over the partitioned NTCB in various ways. Some part, all, or none of the DAC may be implemented in a given component of the network system. In particular, com- ponents that support only internal subjects (i.e., that have no subjects acting as direct surrogates for users), such as a public network packet switch, might not implement the DAC mechanism(s) directly (e.g., they are unlikely to contain access control lists). Identification of users by groups may be achieved in various ways in the networking environment. For example, the network identifiers (e.g., internet addresses) for vari- ous components (e.g., hosts, gateways) can be used as iden- tifiers of groups of individual users (e.g., "all users at Host A," "all users of network Q") so long as the individu- als involved in the group are implied by the group identif- ier. For example, Host A might employ a particular group-id, for which it maintains a list of explicit users in that group, in its network exchange with Host B, which accepts the group-id under the conditions of this interpretation. For networks, individual hosts will impose need-to-know controls over their users on the basis of named individuals - much like (in fact, probably the same) controls used when there is no network connection. When group identifiers are acceptable for access con- trol, the identifier of some other host may be employed, to eliminate the maintenance that would be required if indivi- dual identification of remote users was employed. In class C2 and higher, however, it must be possible from that audit record to identify (immediately or at some later time) exactly the individuals represented by a group identifier at the time of the use of that identifier. There is allowed to be an uncertainty because of elapsed time between changes in the group membership and the enforcement in the access con- trol mechanisms. The DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. The reference monitor manages all the physical resources of the system and from them creates the abstraction of subjects and objects that it con- trols. Some of these subjects and objects may be used to implement a part of the NTCB. When the DAC mechanism is distributed in such NTCB subjects (i.e., when outside the reference monitor), the assurance requirements (see the Assurance section) for the design and implementation of the DAC shall be those of class C2 for all networks of class C2 or above. When integrity is included as part of the network dis- cretionary security policy, the above interpretations shall be specifically applied to the controls over modification, viz, the write mode of access, within each component based on identified users or groups of users. + Rationale In this class, the supporting elements of the overall DAC mechanism are required to isolate information (objects) that supports DAC so that it is subject to auditing require- ments (see the System Architecture section). The use of network identifiers to identify groups of individual users could be implemented, for example, as an X.25 community of interest in the network protocol layer (layer 3). In all other respects, the supporting elements of the overall DAC mechanism are treated exactly as untrusted subjects are treated with respect to DAC in an ADP system, with the same result as noted in the interpretation. A typical situation for DAC is that a surrogate process for a remote user will be created in some host for access to objects under the control of the NTCB partition within that host. The interpretation requires that a user identifier be assigned and maintained for each such process by the NTCB, so that access by a surrogate process is subject to essen- tially the same discretionary controls as access by a pro- cess acting on behalf of a local user would be. However, within this interpretation a range of possible interpreta- tions of the assigned user identification is permitted. The most obvious situation would exist if a global database of network users were to be made permanently avail- able on demand to every host, (i.e., a name server existed) so that all user identifications were globally meaningful. It is also acceptable, however, for some NTCB parti- tions to maintain a database of locally-registered users for its own use. In such a case, one could choose to inhibit the creation of surrogate processes for locally unregistered users, or (if permitted by the local policy) alternatively, to permit the creation of surrogate processes with preselected user and group identifiers which, in effect, identify the process as executing on behalf of a member of a group of users on a particular remote host. The intent of the words concerning audit in the interpretation is to pro- vide a minimally acceptable degree of auditability for cases such as the last described. What is required is that there be a capability, using the audit facilities provided by the network NTCB partitions involved, to determine who was logged in at the actual host of the group of remote users at the time the surrogate processing occured. Associating the proper user id with a surrogate process is the job of identification and authentication. This means that DAC is applied locally, with respect to the user id of the surrogate process. The transmission of the data back across the network to the user's host, and the creation of a copy of the data there, is not the business of DAC. Components that support only internal subjects impact the implementation of the DAC by providing services by which information (e.g., a user-id) is made available to a com- ponent that makes a DAC decision. An example of the latter would be the case that a user at Host A attempts to access a file at Host B. The DAC decision might be (and usually would be) made at Host B on the basis of a user-id transmit- ted from Host A to Host B. Unique user identification may be achieved by a variety of mechanisms, including (a) a requirement for unique iden- tification and authentication on the host where access takes place; (b) recognition of fully qualified network addresses authenticated by another host and forwarded to the host where access takes place; or (c) administrative support of a network-wide unique personnel identifier that could be authenticated and forwarded by another host as in (b) above, or could be authenticated and forwarded by a dedicated net- work identification and authentication server. The proto- cols which implement (b) or (c) are subject to the System Architecture requirements. Network support for DAC might be handled in other ways than that described as "typical" above. In particular, some form of centralized access control is often proposed. An access control center may make all decisions for DAC, or it may share the burden with the hosts by controlling host-to- host connections, and leaving the hosts to decide on access to their objects by users at a limited set of remote hosts. In this case the access control center provides the linkage between the connection oriented abstraction (as discussed in the Introduction) and the overall network security policy for DAC. In all cases the enforcement of the decision must be provided by the host where the object resides. There are two forms of distribution for the DAC mechan- ism: implementing portions of the DAC in separate com- ponents, and supporting the DAC in subjects contained within the NTCB partition in a component. Since "the ADP system" is understood to be "the computer network" as a whole, each network component is responsible for enforcing security in the mechanisms allocated to it to ensure secure implementa- tion of the network security policy. For traditional host systems it is frequently easy to also enforce the DAC along with the MAC within the reference monitor, per se, although a few approaches, such as virtual machine monitors, support DAC outside this interface. In contrast to the universally rigid structure of man- datory policies (see the Mandatory Access Control section), DAC policies tend to be very network and system specific, with features that reflect the natural use of the system. For networks it is common that individual hosts will impose controls over their local users on the basis of named individuals-much like the controls used when there is no network connection. However, it is difficult to manage in a centralized manner all the individuals using a large net- work. Therefore, users on other hosts are commonly grouped together so that the controls required by the network DAC policy are actually based on the identity of the hosts or other components. A gateway is an example of such a com- ponent. The assurance requirements are at the very heart of the concept of a trusted system. It is the assurance that determines if a system or network is appropriate for a given environment, as reflected, for example, in the Environments Guideline-. In the case of monolithic systems that have DAC integral to the reference monitor, the assurance require- ments for DAC are inseparable from those of the rest of the reference monitor. For networks there is typically a much clearer distinction due to distributed DAC. The rationale for making the distinction in this network interpretation is that if major trusted network components can be made signi- ficantly easier to design and implement without reducing the ability to meet security policy, then trusted networks will be more easily available. 3.2.1.2 Object Reuse + Statement from DoD 5200.28-STD All authorizations to the information contained within a storage object shall be revoked prior to initial assignment, allocation or reallocation to a subject from the TCB's pool of unused storage objects. No information, including encrypted representations of information, produced by a prior subject's actions is to be available to any subject that obtains access to an object that has been released back to the system. _________________________ - Guidance for Applying the Department of Defense ________ ___ ________ ___ __________ __ _______ Trusted Computer System Evaluation Criteria in Specific _______ ________ ______ __________ ________ __ ________ Environments, CSC-STD-003-85. ____________ + Interpretation The NTCB shall ensure that any storage objects that it controls (e.g., message buffers under the control of a NTCB partition in a component) contain no information for which a subject in that component is not authorized before granting access. This requirement must be enforced by each of the NTCB partitions. + Rationale In a network system, storage objects of interest are things that the NTCB directly controls, such as message buffers in components. Each component of the network system must enforce the object reuse requirement with respect to the storage objects of interest as determined by the network security policy. For example, the DAC requirement in this division leads to the requirement here that message buffers be under the control of the NTCB partition. A buffer assigned to an internal subject may be reused at the discre- tion of that subject which is responsible for preserving the integrity of message streams. Such controlled objects may be implemented in physical resources, such as buffers, disk sectors, tape space, and main memory, in components such as network switches. 3.2.1.3 Labels + Statement from DoD 5200.28-STD Sensitivity labels associated with each ADP SYSTEM RESOURCE (E.G., SUBJECT, STORAGE OBJECT, ROM) THAT IS DIRECTLY OR INDIRECTLY ACCESSIBLE BY SUBJECTS EXTERNAL TO THE TCB shall be maintained by the TCB. These labels shall be used as the basis for mandatory access control decisions. In order to import non-labeled data, the TCB shall request and receive from an authorized user the sensitivity level of the data, and all such actions shall be auditable by the TCB. + Interpretation Non-labeled data imported under the control of the NTCB partition will be assigned a label constrained by THE DEVICE LABELS OF the single-level device used to import it. Labels may include secrecy and integrity- components in accordance with the overall network security policy described by the network sponsor. Whenever the term "label" is used throughout this interpretation, it is understood to include both components as applicable. Similarly, the terms "single-level" and "multilevel" are understood to be based _________________________ - See, for example, Biba, K.J., "Integrity Considera- tion for Secure Computer Systems," ESD-TR-76-372, MTR- 3153, The MITRE Corporation, Bedford, MA, April 1977. on both the secrecy and integrity components of the policy. The mandatory integrity policy will typically have require- ments, such as the probability of undetected message stream modification, that will be reflected in the label for the data so protected. For example, when data is imported its integrity label may be assigned based on mechanisms, such as cryptography, used to provide the assurance required by the policy. The NTCB shall assure that such mechanism are pro- tected from tampering and are always invoked when they are the basis for a label. IF THE SECURITY POLICY INCLUDES AN INTEGRITY POLICY, ALL ACTIVITIES THAT RESULT IN MESSAGE-STREAM MODIFICATION DURING TRANSMISSION ARE REGARDED AS UNAUTHORIZED ACCESSES IN VIOLATION OF THE INTEGRITY POLICY. THE NTCB SHALL HAVE AN AUTOMATED CAPABILITY FOR TESTING, DETECTING, AND REPORTING THOSE ERRORS/CORRUPTIONS THAT EXCEED SPECIFIED NETWORK INTEGRITY POLICY REQUIREMENTS. MESSAGE-STREAM MODIFICATION (MSM) COUNTERMEASURES SHALL BE IDENTIFIED. A TECHNOLOGY OF ADEQUATE STRENGTH SHALL BE SELECTED TO RESIST MSM. IF ENCRYPTION METHODOLOGIES ARE EMPLOYED, THEY SHALL BE APPROVED BY THE NATIONAL SECURITY AGENCY. ALL OBJECTS MUST BE LABELED WITHIN EACH COMPONENT OF THE NETWORK THAT IS TRUSTED TO MAINTAIN SEPARATION OF MULTI- PLE LEVELS OF INFORMATION. THE LABEL ASSOCIATED WITH ANY OBJECTS ASSOCIATED WITH SINGLE-LEVEL COMPONENTS WILL BE IDENTICAL TO THE LEVEL OF THAT COMPONENT. OBJECTS USED TO STORE NETWORK CONTROL INFORMATION, AND OTHER NETWORK STRUC- TURES, SUCH AS ROUTING TABLES, MUST BE LABELED TO PREVENT UNAUTHORIZED ACCESS AND/OR MODIFICATION. + Rationale The interpretation is an extension of the requirement into the context of a network system and partitioned NTCB as defined for these network interpretations. A single-level device may be regarded either as a subject or an object. A multilevel device is regarded as a trusted subject in which the security range of the subject is the minimum-maximum range of the data expected to be transmitted over the dev- ice. The sensitivity labels for either secrecy or integrity or both may reflect non-hierarchical categories or hierarch- ical classification or both. FOR A NETWORK IT IS NECESSARY THAT THIS REQUIREMENT BE APPLIED TO ALL NETWORK SYSTEM RESOURCES AT THE (B2) LEVEL AND ABOVE. THE NTCB IS RESPONSIBLE FOR IMPLEMENTING THE NETWORK INTEGRITY POLICY, WHEN ONE EXISTS. THE NTCB MUST ENFORCE THAT POLICY BY ENSURING THAT INFORMATION IS ACCURATELY TRANSMITTED FROM SOURCE TO DESTINATION (REGARDLESS OF THE NUMBER OF INTERVENING CONNECTING POINTS). THE NTCB MUST BE ABLE TO COUNTER EQUIPMENT FAILURE, ENVIRONMENTAL DISRUP- TIONS, AND ACTIONS BY PERSONS AND PROCESSES NOT AUTHORIZED TO ALTER THE DATA. PROTOCOLS THAT PERFORM CODE OR FORMAT CONVERSION SHALL PRESERVE THE INTEGRITY OF DATA AND CONTROL INFORMATION. THE PROBABILITY OF AN UNDETECTED TRANSMISSION ERROR MAY BE SPECIFIED AS PART OF THE NETWORK SECURITY POLICY SO THAT THE ACCEPTABILITY OF THE NETWORK FOR ITS INTENDED APPLICA- TION MAY BE DETERMINED. THE SPECIFIC METRICS (E.G., PROBA- BILITY OF UNDETECTED MODIFICATION) SATISFIED BY THE DATA CAN BE REFLECTED IN THE INTEGRITY SENSITIVITY LABEL ASSOCIATED WITH THE DATA WHILE IT IS PROCESSED WITHIN A COMPONENT. IT IS RECOGNIZED THAT DIFFERENT APPLICATIONS AND OPERATIONAL ENVIRONMENTS (E.G., CRISIS AS COMPARED TO LOGISTIC) WILL HAVE DIFFERENT INTEGRITY REQUIREMENTS. THE NETWORK SHALL ALSO HAVE AN AUTOMATED CAPABILITY OF TESTING FOR, DETECTING, AND REPORTING ERRORS THAT EXCEED A THRESHOLD CONSISTENT WITH THE OPERATIONAL MODE REQUIREMENTS. THE EFFECTIVENESS OF INTEGRITY COUNTERMEASURES MUST BE ESTA- BLISHED WITH THE SAME RIGOR AS THE OTHER SECURITY-RELEVANT PROPERTIES SUCH AS SECRECY. CRYPTOGRAPHY IS OFTEN UTILIZED AS A BASIS TO PROVIDE DATA INTEGRITY ASSURANCE. MECHANISMS, SUCH AS MANIPULATION DETECTION CODES (MDC)-, MAY BE USED. THE ADEQUACY OF THE ENCRYPTION OR MDC ALGORITHM, THE CORRECTNESS OF THE PROTOCOL LOGIC, AND THE ADEQUACY OF IMPLEMENTATION MUST BE ESTA- BLISHED IN MSM COUNTERMEASURES DESIGN. 3.2.1.3.1 Label Integrity + Statement from DoD 5200.28-STD Sensitivity labels shall accurately represent sensitivity levels of the specific subjects or objects with which they are associated. When exported by the TCB, sensitivity labels shall accurately and unambiguously represent the internal labels and shall be associated with the information being exported. + Interpretation The phrase "exported by the TCB" is understood to include transmission of information from an object in one component to an object in another component. Information transferred between NTCB partitions is addressed in the Sys- tem Integrity Section. The form of internal and external (exported) sensitivity labels may differ, but the meaning shall be the same. The NTCB shall, in addition, ensure that _________________________ - See Jueneman, R. R., "Electronic Document Authenti- cation," IEEE Network Magazine, April 1987, pp 17-23. ____ _______ ________ correct association of sensitivity labels with the informa- tion being transported across the network is preserved. As mentioned in the Trusted Facility Manual Section, encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity level of the ciphertext is generally lower than the cleartext. It fol- lows that cleartext and ciphertext are contained in dif- ferent objects, each possessing its own label. The label of the cleartext must be preserved and associated with the ciphertext so that it can be restored when the cleartext is subsequently obtained by decrypting the ciphertext. If the cleartext is associated with a single-level device, the label of that cleartext may be implicit. The label may also be implicit in the key. When information is exported to an environment where it is subject to deliberate or accidental modification, the TCB shall support the means, such as cryptographic checksums, to assure the accuracy of the labels. When there is a manda- tory integrity policy, the policy will define the meaning of integrity labels. + Rationale Encryption algorithms and their implementation are out- side the scope of these interpretations. Such algorithms may be implemented in a separate device or may be incor- porated in a subject of a larger component. Without preju- dice, either implementation packaging is referred to as an encryption mechanism herein. If encryption methodologies are employed in this regard, they shall be approved by the National Security Agency (NSA). The encryption process is part of the Network Trusted Computer Base partition in the components in which it is implemented. The encryption mechanism is not necessarily a mul- tilevel device or multilevel subject, as these terms are used in these criteria. The process of encryption is mul- tilevel by definition. The cleartext and ciphertext inter- faces carry information of different sensitivity. An encryption mechanism does not process data in the sense of performing logical or arithmetic operations on that data with the intent of producing new data. The cleartext and ciphertext interfaces on the encryption mechanism must be separately identified as being single-level or multilevel. If the interface is single-level, then the sensitivity of the data is established by a trusted individual and impli- citly associated with the interface; the Exportation to Single-Level Devices criterion applies. If the interface is multilevel, then the data must be labeled; the Exportation to Multilevel Devices criterion applies. The network architect is free to select an accept- able mechanism for associating a label with an object. With reference to encrypted objects, the following examples are possible: 1. Include a label field in the protocol definition of the object. 2. Implicitly associate the label with the object through the encryption key. That is, the encryption key uniquely identifies a sensitivity level. A sin- gle or private key must be protected at the level of the data that it encrypts. 3.2.1.3.2 Exportation of Labeled Information + Statement from DoD 5200.28-STD The TCB shall designate each communication channel and I/O device as either single-level or multilevel. Any change in this designation shall be done manually and shall be audit- able by the TCB. The TCB shall maintain and be able to audit any change in the sensitivity level or levels associ- ated with a communications channel or I/O device. + Interpretation Each communication channel and network component shall be designated as either single-level or multilevel. Any change in this designation shall be done with the cognizance and approval of the administrator or security officer in charge of the affected components and the administrator or security officer in charge of the NTCB. This change shall be auditable by the network. The NTCB shall maintain and be able to audit any change in the DEVICE LABELS associated with a single-level communication channel or the range asso- ciated with a multilevel communication channel or component. The NTCB shall also be able to audit any change in the set of sensitivity levels associated with the information which can be transmitted over a multilevel communication channel or component. + Rationale Communication channels and components in a network are analogous to communication channels and I/O devices in stand-alone systems. They must be designated as either mul- tilevel (i.e., able to distinguish and maintain separation among information of various sensitivity levels) or single- level. As in the TCSEC, single-level devices may only be attached to single-level channels. The level or set of levels of information that can be sent to a component or over a communication channel shall only change with the knowledge and approval of the security officers (or system administrator, if there is no security officer) of the network, and of the affected components. This requirement ensures that no significant security- relevant changes are made without the approval of all affected parties. 3.2.1.3.2.1 Exportation to Multilevel Devices + Statement from DoD 5200.28-STD When the TCB exports an object to a multilevel I/O device, the sensitivity label associated with that object shall also be exported and shall reside on the same physical medium as the exported information and shall be in the same form (i.e., machine-readable or human-readable form). When the TCB exports or imports an object over a multilevel communi- cations channel, the protocol used on that channel shall provide for the unambiguous pairing between the sensitivity labels and the associated information that is sent or received. + Interpretation The components, including hosts, of a network shall be interconnected over "multilevel communication channels," multiple single-level communication channels, or both, when- ever the information is to be protected at more than a sin- gle sensitivity level. The protocol for associating the sensitivity label and the exported information shall provide the only information needed to correctly associate a sensi- tivity level with the exported information transferred over the multilevel channel between the NTCB partitions in indi- vidual components. This protocol definition must specify the representation and semantics of the sensitivity labels (i.e., the machine-readable label must uniquely represent the sensitivity level). The "unambiguous" association of the sensitivity level with the communicated information shall meet the same level of accuracy as that required for any other label within the NTCB, as specified in the criterion for Label Integrity. This may be provided by protected and highly reliable direct physical layer connections, or by traditional cryptographic link protection in which any errors during transmission can be readily detected, or by use of a separate channel. THE RANGE OF INFORMATION IMPORTED OR EXPORTED MUST BE CON- STRAINED BY THE ASSOCIATED DEVICE LABELS. + Rationale This protocol must specify the representation and semantics of the sensitivity labels. See the Mandatory Access Control Policies section in Appendix B. The mul- tilevel device interface to (untrusted) subjects may be implemented either by the interface of the reference moni- tor, per se, or by a multilevel subject (e.g., a "trusted subject" as defined in the Bell-LaPadula Model) that pro- vides the labels based on the internal labels of the NTCB partition. The current state of the art limits the support for mandatory policy that is practical for secure networks. Reference monitor support to ensure the control over all the operations of each subject in the network must be completely provided within the single NTCB partition on which that sub- ject interfaces to the NTCB. This means that the entire portion of the "secure state" represented in the FORMAL security policy model that may be changed by transitions invoked by this subject must be contained in the same com- ponent. The secure state of an NTCB partition may be affected by events external to the component in which the NTCB parti- tion resides (e.g., arrival of a message). The effect occurs asynchronusly after being initiated by an event in another component or partition. For example, indeterminate delays may occur between the initiation of a message in one component, the arrival of the message in the NTCB partition in another component, and the corresponding change to the secure state of the second component. Since each component is executing concurrently, to do otherwise would require some sort of network-wide control to synchronize state tran- sitions, such as a global network-wide clock for all proces- sors; in general, such designs are not practical and prob- ably not even desirable. Therefore, the interaction between NTCB partitions is restricted to just communications between pairs (at least logically) of devices-multilevel devices if the device(s) can send/receive data of more than a single level. For broadcast channels the pairs are the sender and intended receiver(s). However, if the broadcast channel carries multiple levels of information, additional mechanism (e.g., cryptochecksum maintained by the TCB) may be required to enforce separation and proper delivery. A common representation for sensitivity labels is needed in the protocol used on that channel and understood by both the sender and receiver when two multilevel devices (in this case, in two different components) are intercon- nected. Each distinct sensitivity level of the overall net- work policy must be represented uniquely in these labels. Within a monolithic TCB, the accuracy of the sensi- tivity labels is generally assured by simple techniques, e.g., very reliable connections over very short physical connections, such as on a single printed circuit board or over an internal bus. In many network environments there is a much higher probability of accidentally or maliciously introduced errors, and these must be protected against. 3.2.1.3.2.2 Exportation to Single-Level Devices + Statement from DoD 5200.28-STD Single-level I/O devices and single-level communication channels are not required to maintain the sensitivity labels of the information they process. However, the TCB shall include a mechanism by which the TCB and an authorized user reliably communicate to designate the single sensitivity level of information imported or exported via single-level communication channels or I/O devices. + Interpretation Whenever one or both of two directly connected com- ponents is not trusted to maintain the separation of infor- mation of different sensitivity levels, or whenever the two directly connected components have only a single sensitivity level in common, the two components of the network shall communicate over a single-level channel. Single-level com- ponents and single-level communication channels are not required to maintain the sensitivity labels of the informa- tion they process. However, the NTCB shall include a reli- able communication mechanism by which the NTCB and an authorized user (VIA A TRUSTED PATH) or a subject within an NTCB partition can designate the single sensitivity level of information imported or exported via single-level communica- tion channels or network components. THE LEVEL OF INFORMA- TION COMMUNICATED MUST EQUAL THE DEVICE LEVEL. + Rationale Single-level communications channels and single-level components in networks are analogous to single level chan- nels and I/O devices in stand-alone systems in that they are not trusted to maintain the separation of information of different sensitivity levels. The labels associated with data transmitted over those channels and by those components are therefore implicit; the NTCB associates labels with the data because of the channel or component, not because of an explicit part of the bit stream. Note that the sensitivity level of encrypted information is the level of the cipher- text rather than the original level(s) of the plaintext. 3.2.1.3.2.3 Labeling Human-Readable Output + Statement from DoD 5200.28-STD The ADP system administrator shall be able to specify the printable label names associated with exported sensitivity labels. The TCB shall mark the beginning and end of all human-readable, paged, hardcopy output (e.g., line printer output) with human-readable sensitivity labels that prop- erly1 represent the sensitivity of the output. The TCB shall, by default, mark the top and bottom of each page of human-readable, paged, hardcopy output (e.g., line printer output) with human-readable sensitivity labels that prop- erly1 represent the sensitivity of the page. The TCB shall, by default and in an appropriate manner, mark other forms of human readable output (e.g., maps, graphics) with human- readable sensitivity labels that properly1 represent the sensitivity of the output. Any override of these markings defaults shall be auditable by the TCB. _________________________ 1 The hierarchical classification component in human-readable sensitivity labels shall be equal to the greatest hierarchical classification of any of the information in the output that the labels refer to; the non-hierarchical category component shall include all of the non-hierarchical categories of the information in the output the labels refer to, but to no other non- hierarchical categories. + Interpretation This criterion imposes no requirement to a component that produces no human-readable output. For those that do produce human-readable output, each sensitivity level that is defined to the network shall have a uniform meaning across all components. The network administrator, in con- junction with any affected component administrator, shall be able to specify the human-readable label that is associated with each defined sensitivity level. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system and partitioned NTCB as defined for these network interpreta- tions. 3.2.1.3.3 Subject Sensitivity Labels + Statement from DoD 5200.28-STD THE TCB SHALL IMMEDIATELY NOTIFY A TERMINAL USER OF EACH CHANGE IN THE SENSITIVITY LEVEL ASSOCIATED WITH THAT USER DURING AN INTERACTIVE SESSION. A TERMINAL USER SHALL BE ABLE TO QUERY THE TCB AS DESIRED FOR A DISPLAY OF THE SUBJECT'S COMPLETE SENSITIVITY LABEL. + Interpretation AN NTCB PARTITION SHALL IMMEDIATELY NOTIFY A TERMINAL USER ATTACHED TO ITS COMPONENT OF EACH CHANGE IN THE SENSI- TIVITY LEVEL ASSOCIATED WITH THAT USER. + Rationale THE LOCAL NTCB PARTITION MUST ENSURE THAT THE USER UNDERSTANDS THE SENSITIVITY LEVEL OF INFORMATION SENT TO AND FROM A TERMINAL. WHEN A USER HAS A SURROGATE PROCESS IN ANOTHER COMPONENT, ADJUSTMENTS TO ITS LEVEL MAY OCCUR TO MAINTAIN COMMUNICATION WITH THE USER. THESE CHANGES MAY OCCUR ASYNCHRONOUSLY. SUCH ADJUSTMENTS ARE NECESSITATED BY MANDATORY ACCESS CONTROL AS APPLIED TO THE OBJECTS INVOLVED IN THE COMMUNICATION PATH. 3.2.1.3.4 Device Labels + Statement from DoD 5200.28-STD THE TCB SHALL SUPPORT THE ASSIGNMENT OF MINIMUM AND MAXIMUM SENSITIVITY LEVELS TO ALL ATTACHED PHYSICAL DEVICES. THESE SENSITIVITY LEVELS SHALL BE USED BY THE TCB TO ENFORCE CON- STRAINTS IMPOSED BY THE PHYSICAL ENVIRONMENTS IN WHICH THE DEVICES ARE LOCATED. + Interpretation THIS REQUIREMENT APPLIES AS WRITTEN TO EACH NTCB PARTI- TION THAT IS TRUSTED TO SEPARATE INFORMATION BASED ON SENSI- TIVITY LEVEL. EACH I/O DEVICE IN A COMPONENT, USED FOR COM- MUNICATION WITH OTHER NETWORK COMPONENTS, IS ASSIGNED A DEV- ICE RANGE, CONSISTING OF A SET OF LABELS WITH A MAXIMUM AND MINIMUM. (A DEVICE RANGE USUALLY CONTAINS, BUT DOES NOT NECESSARILY CONTAIN, ALL POSSIBLE LABELS "BETWEEN" THE MAX- IMUM AND MINIMUM, IN THE SENSE OF DOMINATING THE MINIMUM AND BEING DOMINATED BY THE MAXIMUM.) THE NTCB ALWAYS PROVIDES AN ACCURATE LABEL FOR INFORMA- TION EXPORTED THROUGH DEVICES. INFORMATION EXPORTED OR IMPORTED USING A SINGLE-LEVEL DEVICE IS LABELLED IMPLICITLY BY THE SENSITIVITY LEVEL OF THE DEVICE. INFORMATION EXPORTED FROM ONE MULTILEVEL DEVICE AND IMPORTED AT ANOTHER MUST BE LABELLED THROUGH AN AGREED-UPON PROTOCOL, UNLESS IT IS LABELLED IMPLICITLY BY USING A COMMUNICATION LINK THAT ALWAYS CARRIES A SINGLE LEVEL. INFORMATION EXPORTED AT A GIVEN SENSITIVITY LEVEL CAN BE SENT ONLY TO AN IMPORTING DEVICE WHOSE DEVICE RANGE CON- TAINS THAT LEVEL OR A HIGHER LEVEL. IF THE IMPORTING DEVICE RANGE DOES NOT CONTAIN THE GIVEN LEVEL, THE INFORMATION IS RELABELLED UPON RECEPTION AT A HIGHER LEVEL WITHIN THE IMPORTING DEVICE RANGE. RELABELLING SHOULD NOT OCCUR OTHER- WISE. + Rationale THE PURPOSE OF DEVICE LABELS IS TO REFLECT AND CON- STRAIN THE SENSITIVITY LEVELS OF INFORMATION AUTHORIZED FOR THE PHYSICAL ENVIRONMENT IN WHICH THE DEVICES ARE LOCATED. THE INFORMATION TRANSFER RESTRICTIONS PERMIT ONE-WAY COMMUNICATION (I.E., NO ACKNOWLEDGEMENTS) FROM ONE DEVICE TO ANOTHER WHOSE RANGES HAVE NO LEVEL IN COMMON, AS LONG AS EACH LEVEL IN THE SENDING DEVICE RANGE IS DOMINATED BY SOME LEVEL IN THE RECEIVING DEVICE RANGE. IT IS NEVER PERMITTED TO SEND INFORMATION AT A GIVEN LEVEL TO A DEVICE WHOSE RANGE DOES NOT CONTAIN A DOMINATING LEVEL. (SEE APPENDIX C FOR SIMILAR INTERCONNECTION RULES FOR THE INTERCONNECTED AIS VIEW.) 3.2.1.4 Mandatory Access Control + Statement from DoD 5200.28-STD The TCB shall enforce a mandatory access control policy over all RESOURCES (I.E., SUBJECTS, STORAGE OBJECTS, AND I/O DEV- ICES) THAT ARE DIRECTLY OR INDIRECTLY ACCESSIBLE BY SUBJECTS EXTERNAL TO THE TCB. These subjects and objects shall be assigned sensitivity labels that are a combination of hierarchical classification levels and non-hierarchical categories, and the labels shall be used as the basis for mandatory access control decisions. The TCB shall be able to support two or more such sensitivity levels. (See the Mandatory Access Control interpretations.) The following requirements shall hold for all accesses between ALL SUB- JECTS EXTERNAL TO THE TCB AND ALL OBJECTS DIRECTLY OR INDIRECTLY ACCESSIBLE BY THESE SUBJECTS. A subject can read an object only if the hierarchical classification in the subject's sensitivity level is greater than or equal to the hierarchical classification of the object's sensitivity level and the non-hierarchical categories in the subject's sensitivity level include all the non-hierarchical categories in the object's sensitivity level. A subject can write an object only if the hierarchical classification in the subject's sensitivity level is less than or equal to the hierarchical classification of the object's sensitivity level and the non-hierarchical categories in the subject's sensitivity level are included in the non-hierarchical categories in the object's sensitivity level. Identification and authentication data shall be used by the TCB to authen- ticate the user's identity and to ensure that the sensi- tivity level and authorization of subjects external to the TCB that may be created to act on behalf of the individual user are dominated by the clearance and authorization of that user. + Interpretation Each partition of the NTCB exercises mandatory access control policy over all subjects and objects in its COM- PONENT. In a network, the responsibility of an NTCB parti- tion encompasses all mandatory access control functions in its component that would be required of a TCB in a stand- alone system. In particular, subjects and objects used for communication with other components are under the control of the NTCB partition. Mandatory access control includes secrecy and integrity control to the extent that the network sponsor has described in the overall network security pol- icy. Conceptual entities associated with communication between two components, such as sessions, connections and virtual circuits, may be thought of as having two ends, one in each component, where each end is represented by a local object. Communication is viewed as an operation that copies information from an object at one end of a communication path to an object at the other end. Transient data-carrying entities, such as datagrams and packets, exist either as information within other objects, or as a pair of objects, one at each end of the communication path. The requirement for "two or more" sensitivity levels can be met by either secrecy or integrity levels. When there is a mandatory integrity policy, the stated require- ments for reading and writing are generalized to: A subject can read an object only if the subject's sensitivity level dominates the object's sensitivity level, and a subject can write an object only if the object's sensitivity level dom- inates the subject's sensitivity level. Based on the integrity policy, the network sponsor shall define the domi- nance relation for the total label, for example, by combin- ing secrecy and integrity lattices. - + Rationale An NTCB partition can maintain access control only over subjects and objects in its component. AT LEVELS B2 AND ABOVE, THE NTCB PARTITION MUST MAINTAIN ACCESS CONTROL OVER ALL SUBJECTS AND OBJECTS IN ITS COMPONENT. Access by a sub- ject in one component to information contained in an object in another component requires the creation of a subject in the remote component which acts as a surrogate for the first subject. The mandatory access controls must be enforced at the interface of the reference monitor (viz. the mechanism that controls physical processing resources) for each NTCB parti- tion. This mechanism creates the abstraction of subjects and objects which it controls. Some of these subjects out- side the reference monitor, per se, may be designated to implement part of an NTCB partition's mandatory policy, e.g., by using the ``trusted subjects" defined in the Bell- LaPadula model. The prior requirements on exportation of labeled infor- mation to and from I/O devices ensure the consistency between the sensitivity labels of objects connected by a communication path. As noted in the introduction, the net- work architecture must recognize the linkage between the overall mandatory network security policy and the connection oriented abstraction. For example, individual data-carrying entities such as datagrams can have individual sensitivity labels that subject them to mandatory access control in each component. The abstraction of a single-level connection is realized and enforced implicitly by an architecture while a connection is realized by single-level subjects that neces- sarily employ only datagrams of the same level. The fundamental trusted systems technology permits the DAC mechanism to be distributed, in contrast to the require- ments for mandatory access control. For networks this separation of MAC and DAC mechanisms is the rule rather than _________________________ - See, for example, Grohn, M. J., A Model of a Pro- _ _____ __ _ ___ tected Data Management System, ESD-TR-76-289, I. P. ______ ____ __________ ______ Sharp Assoc. Ltd., June, 1976; and Denning, D .E., Lunt, T. F., Neumann, P. G., Schell, R. R., Heckman, M. and Shockley, W., Secure Distributed Data Views, Secu- ______ ___________ ____ _____ ____ rity Policy and Interpretation for a Class A1 Multilev- ____ ______ ___ ______________ ___ _ _____ __ ________ el Secure Relational Database System,SRI International, __ ______ __________ ________ ______ November 1986. the exception. The set of total sensitivity labels used to represent all the sensitivity levels for the mandatory access control (combined data secrecy and data integrity) policy always forms a partially ordered set. Without loss of generality, this set of labels can always be extended to form a lattice, by including all the combinations of non-hierarchical categories. As for any lattice, a dominance relation is always defined for the total sensitivity labels. For admin- istrative reasons it may be helpful to have a maximum level which dominates all others. 3.2.2 Accountability _ _ _ ______________ 3.2.2.1 Identification and Authentication + Statement from DoD 5200.28-STD The TCB shall require users to identify themselves to it before beginning to perform any other actions that the TCB is expected to mediate. Furthermore, the TCB shall maintain authentication data that includes information for verifying the identify of individual users (e.g., passwords) as well as information for determining the clearance and authoriza- tions of individual users. This data shall be used by the TCB to authenticate the user's identity and to ensure that the sensitivity level and authorization of subjects external to the TCB that may be created to act on behalf of the indi- vidual user are dominated by the clearance and authorization of that user. The TCB shall protect authentication data so that it cannot be accessed by any unauthorized user. The TCB shall be able to enforce individual accountability by providing the capability to uniquely identify each indivi- dual ADP system user. The TCB shall also provide the capa- bility of associating this identity with all auditable actions taken by that individual. + Interpretation The requirement for identification and authentication of users is the same for a network system as for an ADP sys- tem. The identification and authentication may be done by the component to which the user is directly connected or some other component, such as an identification and authen- tication server. Available techniques, such as those described in the Password Guideline=, are generally also applicable in the network context. However, in cases where the NTCB is expected to mediate actions of a host (or other network component) that is acting on behalf of a user or group of users, the NTCB may employ identification and _________________________ = Department of Defense Password Management Guide- __________ __ _______ ________ __________ _____ line, CSC-STD-002-85 ____ authentication of the host (or other component) in lieu of identification and authentication of an individual user, so long as the component identifier implies a list of specific users uniquely associated with the identifier at the time of its use for authentication. This requirement does not apply to internal subjects. Authentication information, including the identity of a user (once authenticated) may be passed from one component to another without reauthentication, so long as the NTCB protects (e.g., by encryption) the information from unau- thorized disclosure and modification. This protection shall provide at least a similar level of assurance (or strength of mechanism) as pertains to the protection of the authenti- cation mechanism and authentication data. + Rationale The need for accountability is not changed in the con- text of a network system. The fact that the NTCB is parti- tioned over a set of components neither reduces the need nor imposes new requirements. That is, individual accountabil- ity is still the objective. Also, in the context of a net- work system at the (C2) level or higher "individual accoun- tability" can be satisfied by identification of a host (or other component) so long as the requirement for traceability to individual users or a set of specific individual users with active subjects is satisfied. There is allowed to be an uncertainty in traceability because of elapsed time between changes in the group membership and the enforcement in the access control mechanisms. In addition, there is no need in a distributed processing system like a network to reauthen- ticate a user at each point in the network where a projec- tion of a user (via the subject operating on behalf of the user) into another remote subject takes place. The passing of identifiers and/or authentication infor- mation from one component to another is usually done in sup- port to the implementation of the discretionary access con- trol (DAC). This support relates directly to the DAC regarding access by a user to a storage object in a dif- ferent NTCB partition than the one where the user was authenticated. Employing a forwarded identification implies additional reliance on the source and components along the path. If the authenticated identification is used as the basis of determining a sensitivity label for a subject, it must satisfy the Label Integrity criterion. An authenticated identification may be forwarded between components and employed in some component to iden- tify the sensitivity level associated with a subject created to act on behalf of the user so identified. 3.2.2.1.1 Trusted Path + Statement from DoD 5200.28-STD THE TCB SHALL SUPPORT A TRUSTED COMMUNICATION PATH BETWEEN ITSELF AND USER FOR INITIAL LOGIN AND AUTHENTICATION. COM- MUNICATIONS VIA THIS PATH SHALL BE INITIATED EXCLUSIVELY BY A USER. + Interpretation A TRUSTED PATH IS SUPPORTED BETWEEN A USER (I.E., HUMAN) AND THE NTCB PARTITION IN THE COMPONENT TO WHICH THE USER IS DIRECTLY CONNECTED. + Rationale WHEN A USER LOGS INTO A REMOTE COMPONENT, THE USER ID IS TRANSMITTED SECURELY BETWEEN THE LOCAL AND REMOTE NTCB PARTITIONS IN ACCORDANCE WITH THE REQUIREMENTS IN IDENTIFI- CATION AND AUTHENTICATION. TRUSTED PATH IS NECESSARY IN ORDER TO ASSURE THAT THE USER IS COMMUNICATING WITH THE NTCB AND ONLY THE NTCB WHEN SECURITY RELEVANT ACTIVITIES ARE TAKING PLACE (E.G., AUTHEN- TICATE USER, SET CURRENT SESSION SENSITIVITY LEVEL). HOW- EVER, TRUSTED PATH DOES NOT ADDRESS COMMUNICATIONS WITHIN THE NTCB, ONLY COMMUNICATIONS BETWEEN THE USER AND THE NTCB. IF, THEREFORE, A COMPONENT DOES NOT SUPPORT ANY DIRECT USER COMMUNICATION THEN THE COMPONENT NEED NOT CONTAIN MECHANISMS FOR ASSURING DIRECT NTCB TO USER COMMUNICATIONS. THE REQUIREMENT FOR TRUSTED COMMUNICATION BETWEEN ONE NTCB PARTITION AND ANOTHER NCTB PARTITION IS ADDRESSED IN THE SYSTEM ARCHITECTURE SECTION. THESE REQUIREMENTS ARE SEPARATE AND DISTINCT FROM THE USER TO NTCB COMMUNICATION REQUIREMENT OF A TRUSTED PATH. HOWEVER, IT IS EXPECTED THAT THIS TRUSTED COMMUNICATION BETWEEN ONE NTCB PARTITION AND ANOTHER NTCB PARTITION WILL BE USED IN CONJUNCTION WITH THE TRUSTED PATH TO IMPLEMENT TRUSTED COMMUNICATION BETWEEN THE USER AND THE REMOTE NTCB PARTITION. 3.2.2.2 Audit _ _ _ _ _____ + Statement from DoD 5200.28-STD The TCB shall be able to create, maintain, and protect from modification or unauthorized access or destruction an audit trail of accesses to the objects it protects. The audit data shall be protected by the TCB so that read access to it is limited to those who are authorized for audit data. The TCB shall be able to record the following types of events: use of identification and authentication mechanisms, intro- duction of objects into a user's address space (e.g., file open, program initiation), deletion of objects, actions taken by computer operators and system administrators and/or system security officers, and other security relevant events. The TCB shall also be able to audit any override of human-readable output markings. For each recorded event, the audit record shall identify: date and time of the event, user, type of event, and success or failure of the event. For identification/authentication events the origin of request (e.g., terminal ID) shall be included in the audit record. For events that introduce an object into a user's address space and for object deletion events the audit record shall include the name of the object and the object's sensitivity level. The ADP system administrator shall be able to selectively audit the actions of any one or more users based on individual identify and/or object sensitivity level. THE TCB SHALL BE ABLE TO AUDIT THE IDENTIFIED EVENTS THAT MAY BE USED IN THE EXPLOITATION OF COVERT STORAGE CHANNELS. + Interpretation This criterion applies as stated. The sponsor must select which events are auditable. If any such events are not distinguishable by the NTCB alone (for example those identified in Part II), the audit mechanism shall provide an interface, which an authorized subject can invoke with parameters sufficient to produce an audit record. These audit records shall be distinguishable from those provided by the NTCB. In the context of a network system, "other security relevant events" (depending on network system architecture and network security policy) might be as fol- lows: 1. Identification of each access event (e.g., estab- lishing a connection or a connectionless association between processes in two hosts of the network) and its principal parameters (e.g., host identifiers of the two hosts involved in the access event and user identifier or host identifier of the user or host that is requesting the access event) 2. Identification of the starting and ending times of each access event using local time or global syn- chronized time 3. Identification of security-relevant exceptional con- ditions (e.g., potential violation of data integrity, such as misrouted datagrams) detected during the transactions between two hosts 4. Utilization of cryptographic variables 5. Changing the configuration of the network (e.g., a component leaving the network and rejoining) In addition, identification information should be included in appropriate audit trail records, as necessary, to allow association of all related (e.g., involving the same network event) audit trail records (e.g., at different hosts) with each other. Furthermore, a component of the network system may provide the required audit capability (e.g., storage, retrieval, reduction, analysis) for other components that do not internally store audit data but transmit the audit data to some designated collection com- ponent. Provisions shall be made to control the loss of audit data due to unavailability of resources. In the context of a network system, the "user's address space" is extended, for object introduction and deletion events, to include address spaces being employed on behalf of a remote user (or host). However, the focus remains on users in contrast to internal subjects as discussed in the DAC criterion. In addition, audit information must be stored in machine-readable form. THE CAPABILITY MUST EXIST TO AUDIT THE IDENTIFIED EVENTS THAT MAY BE USED IN THE EXPLOITATION OF COVERT STORAGE CHANNELS. TO ACCOMPLISH THIS, EACH NTCB PARTITION MUST BE ABLE TO AUDIT THOSE EVENTS LOCALLY THAT MAY LEAD TO THE EXPLOITATION OF A COVERT STORAGE CHANNEL WHICH EXIST BECAUSE OF THE NETWORK. + Rationale For remote users, the network identifiers (e.g., inter- net address) can be used as identifiers of groups of indivi- dual users (e.g., "all users at Host A") to eliminate the maintenance that would be required if individual identifica- tion of remote users was employed. In this class (C2), how- ever, it must be possible to identify (immediately or at some later time) the individuals represented by a group identifier. In all other respects, the interpretation is a straightforward extension of the criterion into the context of a network system. IDENTIFICATION OF COVERT CHANNEL EVENTS IS ADDRESSED IN THE COVERT CHANNEL ANALYSIS SECTION. 3.2.3 Assurance _ _ _ _________ 3.2.3.1 Operational Assurance 3.2.3.1.1 System Architecture + Statement from DoD 5200.28-STD THE TCB SHALL MAINTAIN A DOMAIN FOR ITS OWN EXECUTION THAT PROTECTS IT FROM EXTERNAL INTERFERENCE OR TAMPERING (E.G., BY MODIFICATION OF ITS CODE OR DATA STRUCTURES). THE TCB SHALL MAINTAIN PROCESS ISOLATION THROUGH THE PROVISION OF DISTINCT ADDRESS SPACES UNDER ITS CONTROL. THE TCB SHALL BE INTERNALLY STRUCTURED INTO WELL-DEFINED LARGELY INDEPENDENT MODULES. IT SHALL MAKE EFFECTIVE USE OF AVAILABLE HARDWARE TO SEPARATE THOSE ELEMENTS THAT ARE PROTECTION-CRITICAL FROM THOSE THAT ARE NOT. THE TCB MODULES SHALL BE DESIGNED SUCH THAT THE PRINCIPLE OF LEAST PRIVILEGE IS ENFORCED. FEATURES IN HARDWARE, SUCH AS SEGMENTATION, SHALL BE USED TO SUPPORT LOGICALLY DISTINCT STORAGE OBJECTS WITH SEPARATE ATTRIBUTES (NAMELY: READABLE, WRITABLE). THE USER INTERFACE TO THE TCB SHALL BE COMPLETELY DEFINED AND ALL ELEMENTS OF THE TCB IDENTIFIED. + Interpretation The system architecture criterion must be met individu- ally by all NTCB partitions. Implementation of the require- ment that the NTCB maintain a domain for its own execution is achieved by having each NTCB partition maintain a domain for its own execution. Since each component is itself a dis- tinct domain in the overall network system, this also satis- fies the requirement for process isolation through distinct address spaces in the special case where a component has only a single subject. THE NTCB MUST BE INTERNALLY STRUCTURED INTO WELL- DEFINED LARGELY INDEPENDENT MODULES AND MEET THE HARDWARE REQUIREMENTS. THIS IS SATISFIED BY HAVING EACH NTCB PARTI- TION SO STRUCTURED. THE NTCB CONTROLS ALL NETWORK RESOURCES. THESE RESOURCES are the union of the sets of resources over which the NTCB partitions have control. Code and data structures belonging to the NTCB, transferred among NTCB subjects (i.e., subjects outside the reference monitor but inside the NTCB) belonging to different NTCB partitions, must be protected against external interference or tamper- ing. For example, a cryptographic checksum or physical means may be employed to protect user authentication data exchanged between NTCB partitions. EACH NTCB PARTITION MUST ENFORCE THE PRINCIPLE OF LEAST PRIVILEGE WITHIN ITS COMPONENT. ADDITIONALLY, THE NTCB MUST BE STRUCTURED SO THAT THE PRINCIPLE OF LEAST PRIVILEGE IS ENFORCED IN THE SYSTEM AS A WHOLE. Each NTCB partition provides isolation of resources (within its component) in accord with the network system architecture and security policy so that "supporting ele- ments" (e.g., DAC and user identification) for the security mechanisms of the network system are strengthened compared to C2, from an assurance point of view, through the provi- sion of distinct address spaces under control of the NTCB. As discussed in the Discretionary Access Control sec- tion, the DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. When distributed in NTCB sub- jects (i.e., when outside the reference monitor), the assurance requirements for the design and implementation of the DAC shall be those of class C2 for all networks of class C2 or above. + Rationale THE REQUIREMENT THAT THE NTCB BE STRUCTURED INTO MODULES AND MEET THE HARDWARE REQUIREMENTS APPLIES WITHIN THE NTCB PARTITIONS IN THE VARIOUS COMPONENTS. THE PRINCIPLE OF LEAST PRIVILEGE REQUIRES THAT EACH USER OR OTHER INDIVIDUAL WITH ACCESS TO THE SYSTEM BE GIVEN ONLY THOSE RESOURCES AND AUTHORIZATIONS REQUIRED FOR THE PERFORMANCE OF THIS JOB. IN ORDER TO ENFORE THIS PRINCIPLE IN THE SYSTEM IT MUST BE ENFORCED IN EVERY NTCB PARTITION THAT SUPPORTS USERS OR OTHER INDIVIDUALS. FOR EXAMPLE, PROHIBITING ACCESS BY ADMINISTRATORS TO OBJECTS OUTSIDE THE NTCB PARTITION (E.G., GAMES) LESSENS THE OPPORTUNITY OF DAM- AGE BY A TROJAN HORSE. The requirement for the protection of communications between NTCB partitions is specifically directed to subjects that are part of the NTCB partitions. Any requirements for such protection for the subjects that are outside the NTCB partitions are addressed in response to the integrity requirements of the security policy. The provision of distinct address spaces under the con- trol of the NTCB provides the ability to separate subjects according to sensitivity level. This requirement is intro- duced at B1 since it is an absolute necessity in order to implement mandatory access controls. 3.2.3.1.2 System Integrity + Statement from DoD 5200.28-STD Hardware and/or software features shall be provided that can be used to periodically validate the correct operation of the on-site hardware and firmware elements of the TCB. + Interpretation Implementation of the requirement is partly achieved by having hardware and/or software features that can be used to periodically validate the correct operation of the hardware and firmware elements of each component's NTCB partition. Features shall also be provided to validate the identity and correct operation of a component prior to its incorporation in the network system and throughout system operation. For example, a protocol could be designed that enables the com- ponents of the partitioned NTCB to exchange messages period- ically and validate each other's correct response. The pro- tocol shall be able to determine the remote entity's ability to respond. NTCB partitions shall provide the capability to report to network administrative personnel the failures detected in other NTCB partitions. Intercomponent protocols implemented within a NTCB shall be designed in such a way as to provide correct opera- tion in the case of failures of network communications or individual components. The allocation of mandatory and dis- cretionary access control policy in a network may require communication between trusted subjects that are part of the NTCB partitions in different components. This communication is normally implemented with a protocol between the subjects as peer entities. Incorrect access within a component shall not result from failure of an NTCB partition to communicate with other components. + Rationale The first paragraph of the interpretation is a straightforward extension of the requirement into the con- text of a network system and partitioned NTCB as defined for these network criteria. NTCB protocols should be robust enough so that they permit the system to operate correctly in the case of local- ized failure. The purpose of this protection is to preserve the integrity of the NTCB itself. It is not unusual for one or more components in a network to be inoperative at any time, so it is important to minimize the effects of such failures on the rest of the network. Additional integrity and denial of service issues are addressed in Part II. 3.2.3.1.3 Covert Channel Analysis + Statement from DoD 5200.28-STD THE SYSTEM DEVELOPER SHALL CONDUCT A THOROUGH SEARCH FOR COVERT STORAGE CHANNELS AND MAKE A DETERMINATION (EITHER BY ACTUAL MEASUREMENT OR BY ENGINEERING ESTIMATION) OF THE MAX- IMUM BANDWIDTH OF EACH IDENTIFIED CHANNEL. (SEE THE COVERT CHANNELS GUIDELINE SECTION.) + Interpretation THE REQUIREMENT, INCLUDING THE TCSEC COVERT CHANNEL GUIDELINE, APPLIES AS WRITTEN. IN A NETWORK, THERE ARE ADDITIONAL INSTANCES OF COVERT CHANNELS ASSOCIATED WITH COM- MUNICATION BETWEEN COMPONENTS. + Rationale THE EXPLOITATION OF NETWORK PROTOCOL INFORMATION (E.G., HEADERS) CAN RESULT IN COVERT STORAGE CHANNELS. THE TOPIC HAS BEEN ADDRESSED IN THE LITERATURE.- _________________________ - See, for example, Girling, C. G., "Covert Channels in LAN's," IEEE Transactions on Software Engineering, ____ ____________ __ ________ ___________ Vol. SE-13, No. 2, February 1987; and Padlipsky, M. A., Snow, D. P., and Karger, P. A., Limitations of End-to- ___________ __ ___ __ End Encryption in Secure Computer Networks, MITRE ___ __________ __ ______ ________ ________ Technical Report, MTR-3592, Vol. I, May 1978 (ESD TR 3.2.3.1.4 Trusted Facility Management + Statement from DoD 5200.28-STD THE TCB SHALL SUPPORT SEPARATE OPERATOR AND ADMINISTRATOR FUNCTIONS. + Interpretation THIS REQUIREMENT APPLIES AS WRITTEN TO BOTH THE NETWORK AS A WHOLE AND TO INDIVIDUAL COMPONENTS WHICH SUPPORT SUCH PERSONNEL. + Rationale IT IS RECOGNIZED THAT BASED ON THE ALLOCATED POLICY ELEMENTS SOME COMPONENTS MAY OPERATE WITH NO HUMAN INTER- FACE. 3.2.3.2 Life-Cycle Assurance 3.2.3.2.1 Security Testing + Statement from DoD 5200.28-STD The security mechanisms of the ADP system shall be tested and found to work as claimed in the system documentation. A team of individuals who thoroughly understand the specific implementation of the TCB shall subject its design documen- tation, source code, and object code to through analysis and testing. Their objectives shall be: to uncover all design and implementation flaws that would permit a subject exter- nal to the TCB to read, change, or delete data normally denied under the mandatory or discretionary security policy enforced by the TCB; as well as to assure that no subject (without authorization to do so) is able to cause the TCB to enter a state such that it is unable to respond to communi- cations initiated by other users. THE TCB SHALL BE FOUND RELATIVELY RESISTANT TO PENETRATION. All discovered flaws shall be removed or neutralized and the TCB retested to demonstrate that they have been eliminated and that new flaws have not been introduced. TESTING SHALL DEMONSTRATE THAT THE TCB IMPLEMENTATION IS CONSISTENT WITH THE DESCRIP- TIVE TOP-LEVEL SPECIFICATION. (See the Security Testing Guidelines.) + Interpretation Testing of a component will require a testbed that exercises the interfaces and protocols of the component including tests under exceptional conditions. The testing of a security mechanism of the network system for meeting this criterion shall be an integrated testing procedure involving all components containing an NTCB partition that implement the given mechanism. This integrated testing is additional to any individual component tests involved in the evaluation of the network system. The sponsor should iden- tify the allowable set of configurations including the sizes of the networks. Analysis or testing procedures and tools shall be available to test the limits of these configura- tions. A change in configuration within the allowable set of configurations does not require retesting. The testing of each component will include the intro- duction of subjects external to the NTCB partition for the component that will attempt to read, change, or delete data normally denied. If the normal interface to the component does not provide a means to create the subjects needed to conduct such a test, then this portion of the testing shall use a special version of the untrusted software for the com- ponent that results in subjects that make such attempts. The results shall be saved for test analysis. Such special versions shall have an NTCB partition that is identical to that for the normal configuration of the component under evaluation. The testing of the mandatory controls shall include tests to demonstrate that the labels for information imported and/or exported to/from the component accurately represent the labels maintained by the NTCB partition for the component for use as the basis for its mandatory access control decisions. The tests shall include each type of device, whether single-level or multilevel, supported by the component. THE NTCB MUST BE FOUND RELATIVELY RESISTANT TO PENETRA- TION. THIS APPLIES TO THE NTCB AS A WHOLE, AND TO EACH NTCB PARTITION IN A COMPONENT OF THIS CLASS. + Rationale The phrase "no subject (without authorization to do so) is able to cause the TCB to enter a state such that it is unable to respond to communications initiated by other users" relates to the security services (Part II of this TNI) for the Denial of Service problem, and to correctness of the protocol implementations. Testing is an important method available in this evaluation division to gain any assurance that the security mechanisms perform their intended function. A major purpose of testing is to demonstrate the system's response to inputs to the NTCB partition from untrusted (and possibly mali- cious) subjects. In contrast to general purpose systems that allow for the dynamic creation of new programs and the introductions of new processes (and hence new subjects) with user speci- fied security properities, many network components have no method for introducing new programs and/or processes during their normal operation. Therefore, the programs necessary for the testing must be introduced as special versions of the software rather than as the result of normal inputs by the test team. However, it must be insured that the NTCB partition used for such tests is identical to the one under evaluation. Sensitivity labels serve a critical role in maintaining the security of the mandatory access controls in the net- work. Especially important to network security is the role of the labels for information communicated between com- ponents - explicit labels for multilevel devices and impli- cit labels for single-level devices. Therefore the testing for correct labels is highlighted. THE REQUIREMENT FOR TESTING TO DEMONSTRATE CONSISTENCY BETWEEN THE NTCB IMPLEMENTATION AND THE DTLS IS A STRAIGHT- FORWARD EXTENSION OF THE TCSEC REQUIREMENT INTO THE CONTEXT OF A NETWORK SYSTEM. 3.2.3.2.2 Design Specification and Verification + Statement from DoD 5200.28-STD A FORMAL model of the security policy supported by the TCB shall be maintained over the life cycle of the ADP system THAT IS PROVEN and demonstrated to be consistent with its axioms. A DESCRIPTIVE TOP-LEVEL SPECIFICATION (DTLS) OF THE TCB SHALL BE MAINTAINED THAT COMPLETELY AND ACCURATELY DESCRIBES THE TCB IN TERMS OF EXCEPTIONS, ERROR MESSAGES, AND EFFECTS. IT SHALL BE SHOWN TO BE AN ACCURATE DESCRIP- TION OF THE TCB INTERFACE. + Interpretation The overall network security policy expressed in this model will provide the basis for the mandatory access con- trol policy exercised by the NTCB over subjects and storage objects in the entire network. The policy will also be the basis for the discretionary access control policy exercised by the NTCB to control access of named users to named objects. Data integrity requirements addressing the effects of unauthorized MSM need not be included in this model. The overall network policy must be decomposed into policy ele- ments that are allocated to appropriate components and used as the basis for the security policy model for those com- ponents. The level of abstraction of the model, and the set of subjects and objects that are explicitly represented in the model, will be affected by the NTCB partitioning. Subjects and objects must be represented explicitly in the model for the partition if there is some network component whose NTCB partition exercises access control over them. The model shall be structured so that the axioms and entities applicable to individual network components are manifest. Global network policy elements that are allocated to com- ponents shall be represented by the model for that com- ponent. THE REQUIREMENTS FOR A NETWORK DTLS ARE GIVEN IN THE DESIGN DOCUMENTATION SECTION. + Rationale The treatment of the model depends to a great extent on the degree of integration of the communications service into a distributed system. In a closely coupled distributed sys- tem, one might use a model that closely resembles one appropriate for a stand-alone computer system. In other cases, the model of each partition will be expected to show the role of the NTCB partition in each kind of component. It will most likely clarify the model, although not part of the model, to show access restrictions implied by the system design; for example, subjects representing protocol entities might have access only to objects containing data units at the same layer of protocol. The allocation of subjects and objects to different proto- col layers is a protocol design choice which need not be reflected in the security policy model. 3.2.3.2.3 Configuration Management + Statement from DoD 5200.28-STD DURING DEVELOPMENT AND MAINTENANCE OF THE TCB, A CONFIGURA- TION MANAGEMENT SYSTEM SHALL BE IN PLACE THAT MAINTAINS CON- TROL OF CHANGES TO THE DESCRIPTIVE TOP-LEVEL SPECIFICATION, OTHER DESIGN DATA, IMPLEMENTATION DOCUMENTATION, SOURCE CODE, THE RUNNING VERSION OF THE OBJECT CODE, AND TEST FIX- TURES AND DOCUMENTATION. THE CONFIGURATION MANAGEMENT SYS- TEM SHALL ASSURE A CONSISTENT MAPPING AMONG ALL DOCUMENTA- TION AND CODE ASSOCIATED WITH THE CURRENT VERSION OF THE TCB. TOOLS SHALL BE PROVIDED FOR GENERATION OF A NEW VER- SION OF THE TCB FROM SOURCE CODE. ALSO AVAILABLE SHALL BE TOOLS FOR COMPARING A NEWLY GENERATED VERSION WITH THE PRE- VIOUS TCB VERSION IN ORDER TO ASCERTAIN THAT ONLY THE INTENDED CHANGES HAVE BEEN MADE IN THE CODE THAT WILL ACTU- ALLY BE USED AS THE NEW VERSION OF THE TCB. + Interpretation THE REQUIREMENT APPLIES AS WRITTEN, WITH THE FOLLOWING EXTENSIONS: 1. A CONFIGURATION MANAGEMENT SYSTEM MUST BE IN PLACE FOR EACH NTCB PARTITION. 2. A CONFIGURATION MANAGEMENT PLAN MUST EXIST FOR THE ENTIRE SYSTEM. IF THE CONFIGURATION MANAGEMENT SYS- TEM IS MADE UP OF THE CONGLOMERATION OF THE CONFI- GURATION MANAGEMENT SYSTEMS OF THE VARIOUS NTCB PAR- TITIONS, THEN THE CONFIGURATION MANAGEMENT PLAN MUST ADDRESS THE ISSUE OF HOW CONFIGURATION CONTROL IS APPLIED TO THE SYSTEM AS A WHOLE. + Rationale EACH NTCB PARTITION MUST HAVE A CONFIGURATION MANAGE- MENT SYSTEM IN PLACE, OR ELSE THERE WILL BE NO WAY FOR THE NTCB AS A WHOLE TO HAVE AN EFFECTIVE CONFIGURATION MANAGE- MENT SYSTEM. THE OTHER EXTENSIONS ARE MERELY REFLECTIONS OF THE WAY THAT NETWORKS OPERATE IN PRACTICE. 3.2.4 Documentation. _ _ _ _____________ 3.2.4.1 Security Features User's Guide + Statement from DoD 5200.28-STD A single summary, chapter, or manual in user documentation shall describe the protection mechanisms provided by the TCB, interpretations on their use, and how they interact with one another. + Interpretation This user documentation describes user visible protec- tion mechanisms at the global (network system) level and at the user interface of each component, and the interaction among these. + Rationale The interpretation is an extension of the requirement into the context of a network system as defined for these network criteria. Documentation of protection mechanisms provided by individual components is required by the cri- teria for trusted computer systems that are applied as appropriate for the individual components. 3.2.4.2 Trusted Facility Manual + Statement from DoD 5200.28-STD A manual addressed to the ADP system administrator shall present cautions about functions and privileges that should be controlled when running a secure facility. The procedures for examining and maintaining the audit files as well as the detailed audit record structure for each type of audit event shall be given. The manual shall describe the operator and administrator functions related to security, to include changing the security characteristics of a user. It shall provide interpretations on the consistent and effective use of the protection features of the system, how they interact, how to securely generate a new TCB, and facility procedures, warnings, and privileges that need to be controlled in order to operate the facility in a secure manner. THE TCB MODULES THAT CONTAIN THE REFERENCE VALIDATION MECHANISM SHALL BE IDENTIFIED. THE PROCEDURES FOR SECURE GENERATION OF A NEW TCB FROM SOURCE AFTER MODIFICATION OF ANY MODULES IN THE TCB SHALL BE DESCRIBED. + Interpretation This manual shall contain specifications and procedures to assist the system administrator(s) maintain cognizance of the network configuration. These specifications and pro- cedures shall address the following: 1. The hardware configuration of the network itself; 2. The implications of attaching new components to the network; 3. The case where certain components may periodically leave the network (e.g., by crashing, or by being disconnected) and then rejoin; 4. Network configuration aspects that can impact the security of the network system; (For example, the manual should describe for the network system administrator the interconnections among components that are consistent with the overall network system architecture.) 5. Loading or modifying NTCB software or firmware (e.g., down-line loading). 6. INCREMENTAL UPDATES; THAT IS, IT MUST EXPLICITLY INDICATE WHICH COMPONENTS OF THE NETWORK MAY CHANGE WITHOUT OTHERS ALSO CHANGING. The physical and administrative environmental controls shall be specified. Any assumptions about security of a given network should be clearly stated (e.g., the fact that all communications links must be physically protected to a certain level). THE COMPONENTS OF THE NETWORK THAT FORM THE NTCB MUST BE IDENTIFIED. FURTHERMORE, THE MODULES WITHIN AN NTCB PAR- TITION THAT CONTAIN THE REFERENCE VALIDATION MECHANISM (IF ANY) WITHIN THAT PARTITION MUST BE IDENTIFIED. THE PROCEDURES FOR THE SECURE GENERATION OF A NEW VER- SION (OR COPY) OF EACH NTCB PARTITION FROM SOURCE MUST BE DESCRIBED. THE PROCEDURES AND REQUIREMENTS FOR THE SECURE GENERATION OF THE NTCB NECESSITATED BY CHANGES IN THE NET- WORK CONFIGURATION SHALL BE DESCRIBED. + Rationale There may be multiple system administrators with diverse responsibilities. The technical security measures described by these criteria must be used in conjunction with other forms of security in order to achieve security of the network. Additional forms include administrative security, physical security, emanations security, etc. Extension of this criterion to cover configuration aspects of the network is needed because, for example, proper interconnection of components is typically essential to achieve a correct realization of the network architec- ture. As mentioned in the section on Label Integrity, cryp- tography is one common mechanism employed to protect commun- ication circuits. Encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity of the ciphertext is generally lower than the cleartext. If encryption methodologies are employed, they shall be approved by the National Security Agency (NSA). The encryption algorithm and its implementation are outside the scope of these interpretations. This algorithm and implementation may be implemented in a separate device or may be a function of a subject in a component not dedi- cated to encryption. Without prejudice, either implementa- tion packaging is referred to as an encryption mechanism herein. THE REQUIREMENTS FOR DESCRIPTIONS OF NTCB GENERATION AND IDENTIFICATION OF MODULES AND COMPONENTS THAT FORM THE NTCB ARE STRAIGHTFORWARD EXTENSIONS OF THE TCSEC REQUIRE- MENTS INTO THE NETWORK CONTEXT. IN THOSE CASES WHERE THE VENDOR DOES NOT PROVIDE SOURCE CODE, AN ACCEPTABLE PROCEDURE SHALL BE TO REQUEST THE VENDOR TO PERFORM THE SECURE GENERA- TION. 3.2.4.3 Test Documentation + Statement from DoD 5200.28-STD The system developer shall provide to the evaluators a docu- ment that describes the test plan, test procedures that show how the security mechanisms were tested, and results of the security mechanisms' functional testing. IT SHALL INCLUDE RESULTS OF TESTING THE EFFECTIVENESS OF THE METHODS USED TO REDUCE COVERT CHANNEL BANDWIDTHS. + Interpretation The "system developer" is interpreted as "the network system sponsor". The description of the test plan should establish the context in which the testing was or should be conducted. The description should identify any additional test components that are not part of the system being evaluated. This includes a description of the test-relevant functions of such test components and a description of the interfacing of those test components to the system being evaluated. The description of the test plan should also demonstrate that the tests adequately cover the network security policy. The tests should include the features described in the System Architecture and the System Integrity sections. The tests should also include network configuration and sizing. + Rationale The entity being evaluated may be a networking subsys- tem (see Appendix A) to which other components must be added to make a complete network system. In that case, this interpretation is extended to include contextual definition because, at evaluation time, it is not possible to validate the test plans without the description of the context for testing the networking subsystem. THE BANDWIDTHS OF COVERT CHANNELS ARE USED TO DETERMINE THE SUITABILITY OF A NETWORK SYSTEM FOR A GIVEN ENVIRONMENT. THE EFFECTIVENESS OF THE METHODS USED TO REDUCE THESE BANDWIDTHS MUST THEREFORE BE ACCURATELY DETERMINED. 3.2.4.4 Design Documentation + Statement from DoD 5200.28-STD Documentation shall be available that provides a description of the manufacturer's philosophy of protection and an expla- nation of how this philosophy is translated into the TCB. THE interfaces between THE TCB modules shall be described. A FORMAL description of the security policy model enforced by the TCB shall be available and an explanation provided to show that it is sufficient to enforce the security policy. The specific TCB protection mechanisms shall be identified and an explanation given to show that they satisfy the model. THE DESCRIPTIVE TOP-LEVEL SPECIFICATION (DTLS) SHALL BE SHOWN TO BE AN ACCURATE DESCRIPTION OF THE TCB INTERFACE. DOCUMENTATION SHALL DESCRIBE HOW THE TCB IMPLEMENTS THE REFERENCE MONITOR CONCEPT AND GIVE AN EXPLANATION WHY IT IS TAMPER RESISTANT, CANNOT BE BYPASSED, AND IS CORRECTLY IMPLEMENTED. DOCUMENTATION SHALL DESCRIBE HOW THE TCB IS STRUCTURED TO FACILITATE TESTING AND TO ENFORCE LEAST PRIVILEGE. THIS DOCUMENTATION SHALL ALSO PRESENT THE RESULTS OF THE COVERT CHANNEL ANALYSIS AND THE TRADEOFFS INVOLVED IN RESTRICTING THE CHANNELS. ALL AUDITABLE EVENTS THAT MAY BE USED IN THE EXPLOITATION OF KNOWN COVERT STORAGE CHANNELS SHALL BE IDENTIFIED. THE BANDWIDTHS OF KNOWN COVERT STORAGE CHANNELS, THE USE OF WHICH IS NOT DETECTABLE BY THE AUDITING MECHANISMS, SHALL BE PROVIDED. (SEE THE COVERT CHANNEL GUIDELINE SECTION.) + Interpretation Explanation of how the sponsor's philosophy of protec- tion is translated into the NTCB shall include a description of how the NTCB is partitioned. The security policy also shall be stated. The description of the interfaces between the NTCB modules shall include the interface(s) between NTCB partitions and modules within the partitions if the modules exist. The sponsor shall describe the security architecture and design, including the allocation of security require- ments among components. THE DOCUMENTATION INCLUDES BOTH A SYSTEM DESCRIPTION AND A SET OF COMPONENT DTLS'S. THE SYSTEM DESCRIPTION ADDRESSES THE NETWORK SECURITY ARCHITECTURE AND DESIGN BY SPECIFYING THE TYPES OF COMPONENTS IN THE NETWORK, WHICH ONES ARE TRUSTED, AND IN WHAT WAY THEY MUST COOPERATE TO SUPPORT NETWORK SECURITY OBJECTIVES. A COMPONENT DTLS SHALL BE PROVIDED FOR EACH TRUSTED NETWORK COMPONENT, I.E., EACH COMPONENT CONTAINING AN NTCB PARTITION. EACH COMPONENT DTLS SHALL DESCRIBE THE INTERFACE TO THE NTCB PARTITION OF ITS COMPONENT. APPENDIX A ADDRESSES COMPONENT EVALUATION ISSUES. As stated in the introduction to Division B, the spon- sor must demonstrate that the NTCB employs the reference monitor concept. The security policy model must be a model for a reference monitor. The security policy model for each partition implement- ing a reference monitor shall fully represent the access control policy supported by the partition, including the discretionary and mandatory security policy for secrecy and/or integrity. For the mandatory policy the single domi- nance relation for sensitivity labels, including secrecy and/or integrity components, shall be precisely defined. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system as defined for this network interpretation. Other documenta- tion, such as description of components and description of operating environment(s) in which the networking subsystem or network system is designed to function, is required else- where, e.g., in the Trusted Facility Manual. In order to be evaluated, a network must possess a coherent Network Security Architecture and Design. (Inter- connection of components that do not adhere to such a single coherent Network Security Architecture is addressed in the Interconnection of Accredited AIS, Appendix C.) The Network Security Architecture must address the security-relevant policies, objectives, and protocols. The Network Security Design specifies the interfaces and services that must be incorporated into the network so that it can be evaluated as a trusted entity. There may be multiple designs that con- form to the same architecture but are more or less incompa- tible and non-interoperable (except through the Interconnec- tion Rules). Security related mechanisms requiring coopera- tion among components are specified in the design in terms of their visible interfaces; mechanisms having no visible interfaces are not specified in this document but are left as implementation decisions. The Network Security Architecture and Design must be available from the network sponsor before evaluation of the network, or any component, can be undertaken. The Network Security Architecture and Design must be sufficiently com- plete, unambiguous, and free from obvious flaws to permit the construction or assembly of a trusted network based on the structure it specifies. When a component is being designed or presented for evaluation, or when a network assembled from components is assembled or presented for evaluation, there must be a priori evidence that the Network security Architecture and Design are satisfied. That is, the components can be assem- bled into a network that conforms in every way with the Net- work Security Architecture and Design to produce a physical realization that is trusted to the extent that its evalua- tion indicates. In order for a trusted network to be constructed from components that can be built independently, the Network Security Architecture and Design must completely and unambi- guously define the security functionality of components as well as the interfaces between or among components. The Network Security Architecture and Design must be evaluated to determine that a network constructed to its specifica- tions will in fact be trusted, that is, it will be evaluat- able under these interpretations. The term "model" is used in several different ways in a network context, e.g., a "protocol reference model," a "for- mal network model," etc. Only the "security policy model" is addressed by this requirement and is specifically intended to model the interface, viz., "security perimeter," of the reference monitor and must meet all the requirements defined in the TCSEC. It must be shown that all parts of the TCB are a valid interpretation of the security policy model, i.e., that there is no change to the secure state except as represented by the model. 3.3 CLASS (B3): SECURITY DOMAINS _ _ _____ __ ________ _______ THE CLASS (B3) NTCB MUST SATISFY THE REFERENCE MONITOR REQUIREMENTS THAT IT MEDIATE ALL ACCESSES OF SUBJECTS TO OBJECTS, BE TAMPERPROOF, AND BE SMALL ENOUGH TO BE SUBJECTED TO ANALYSIS AND TESTS. TO THIS END, THE NTCB IS STRUCTURED TO EXCLUDE CODE NOT ESSENTIAL TO SECURITY POLICY ENFORCEMENT, WITH SIGNIFICANT SYSTEM ENGINEERING DURING NTCB DESIGN AND IMPLEMENTATION DIRECTED TOWARD MINIMIZING ITS COMPLEXITY. A SECURITY ADMINISTRATOR IS SUPPORTED, AUDIT MECHANISMS ARE EXPANDED TO SIGNAL SECURITY-RELEVANT EVENTS, AND SYSTEM RECOVERY PROCEDURES ARE REQUIRED. THE SYS- TEM IS HIGHLY RESISTANT TO PENETRATION. THE FOL- LOWING ARE MINIMAL REQUIREMENTS FOR SYSTEMS ASSIGNED A CLASS (B3) RATING: 3.3.1 Security Policy _ _ _ ________ ______ + Statement from DoD 5200.28-STD Implied from the Introduction to the TCSEC. + Interpretation The network sponsor shall describe the overall network security policy enforced by the NTCB. At a minimum, this policy shall include the discretionary and mandatory requirements applicable to this class. The policy may require data secrecy, or data integrity, or both. The pol- icy is an access control policy having two primary com- ponents: mandatory and discretionary. The policy shall include a discretionary policy for protecting the informa- tion being processed based on the authorizations of indivi- duals, users, or groups of users. This access control pol- icy statement shall describe the requirements on the network to prevent or detect "reading or destroying" sensitive information by unauthorized users or errors. The mandatory policy must define the set of distinct sensitivity levels that it supports. For the Class B1 or above the mandatory policy shall be based on the labels associated with the information that reflects its sensitivity with respect to secrecy and/or integrity, where applicable, and labels asso- ciated with users to reflect their authorization to access such information. Unauthorized users include both those that are not authorized to use the network at all (e.g., a user attempting to use a passive or active wire tap) or a legitimate user of the network who is not authorized to access a specific piece of information being protected. Note that "users" does not include "operators," "system programmers," "technical control officers," "system security officers," and other system support personnel. They are distinct from users and are subject to the Trusted Facility Manual and the System Architecture requirements. Such indi- viduals may change the system parameters of the network sys- tem, for example, by defining membership of a group. These individuals may also have the separate role of users. SECRECY POLICY: The network sponsor shall define the form of the discretionary and mandatory secrecy policy that is enforced in the network to prevent unauthorized users from reading the sensitive infor- mation entrusted to the network. DATA INTEGRITY POLICY: The network sponsor shall define the discretionary and mandatory integrity policy to prevent unauthorized users from modifying, viz., writing, sensitive information. The defini- tion of data integrity presented by the network sponsor refers to the requirement that the informa- tion has not been subjected to unauthorized modifi- cation in the network. The mandatory integrity pol- icy enforced by the NTCB cannot, in general, prevent modification while information is being transmitted between components. However, an integrity sensi- tivity label may reflect the confidence that the information has not been subjected to transmission errors because of the protection afforded during transmission. This requirement is distinct from the requirement for label integrity. + Rationale The word "sponsor" is used in place of alternatives (such as "vendor," "architect," "manufacturer," and "developer") because the alternatives indicate people who may not be available, involved, or relevant at the time that a network system is proposed for evaluation. A trusted network is able to control both the reading and writing of shared sensitive information. Control of writing is used to protect against destruction of informa- tion. A network normally is expected to have policy require- ments to protect both the secrecy and integrity of the information entrusted to it. In a network the integrity is frequently as important or more important than the secrecy requirements. Therefore the secrecy and/or integrity policy to be enforced by the network must be stated for each net- work regardless of its evaluation class. The assurance that the policy is faithfully enforced is reflected in the evaluation class of the network. This control over modification is typically used to protect information so that it may be relied upon and to control the potential harm that would result if the informa- tion were corrupted. The overall network policy require- ments for integrity includes the protection for data both while being processed in a component and while being transmitted in the network. The access control policy enforced by the NTCB relates to the access of subjects to objects within each component. Communications integrity addressed within Part II relates to information while being transmitted. The mandatory integrity policy (at class B1 and above) in some architectures may be useful in supporting the link- age between the connection oriented abstraction introduced in the Introduction and the individual components of the network. For example, in a key distribution center for end-to-end encryption, a distinct integrity category may be assigned to isolate the key generation code and data from possible modification by other supporting processes in the same component, such as operator interfaces and audit. The mandatory integrity policy for some architecture may define an integrity sensitivity label that reflects the specific requirements for ensuring that information has not been subject to random errors in excess of a stated limit nor to unauthorized message stream modification (MSM) -. The specific metric associated with an integrity sensitivity label will generally reflect the intended applications of the network. 3.3.1.1 Discretionary Access Control + Statement from DoD 5200.28-STD The TCB shall define and control access between named users and named objects (e.g., files and programs) in the ADP sys- tem. The enforcement mechanism (e.g., ACCESS CONTROL LISTS) shall allow users to specify and control sharing of those OBJECTS and shall provide controls to limit propagation of access rights. The discretionary access control mechanism shall, either by explicit user action or by default, provide that objects are protected from unauthorized access. These access controls shall be capable of SPECIFYING, FOR EACH _________________________ - See Voydock, Victor L. and Stephen T. Kent, "Secu- rity Mechanisms in High-Level Network Protocols," Com- ___ puting Surveys, Vol. 15, No. 2, June 1983, pp 135-171. ______ _______ NAMED OBJECT, A LIST OF NAMED INDIVIDUALS AND A LIST OF GROUPS OF NAMED INDIVIDUALS WITH THEIR RESPECTIVE MODES OF ACCESS TO THAT OBJECT. FURTHERMORE, FOR EACH SUCH NAMED OBJECT, IT SHALL BE POSSIBLE TO SPECIFY A LIST OF NAMED INDIVIDUALS AND A LIST OF GROUPS OF NAMED INDIVIDUALS FOR WHICH NO ACCESS TO THE OBJECT IS GIVEN. Access permission to an object by users not already possessing access permis- sion shall only be assigned by authorized users. + Interpretation The discretionary access control (DAC) mechanism(s) may be distributed over the partitioned NTCB in various ways. Some part, all, or none of the DAC may be implemented in a given component of the network system. In particular, com- ponents that support only internal subjects (i.e., that have no subjects acting as direct surrogates for users), such as a public network packet switch, might not implement the DAC mechanism(s) directly (e.g., they are unlikely to contain access control lists). Identification of users by groups may be achieved in various ways in the networking environment. For example, the network identifiers (e.g., internet addresses) for vari- ous components (e.g., hosts, gateways) can be used as iden- tifiers of groups of individual users (e.g., "all users at Host A," "all users of network Q") so long as the individu- als involved in the group are implied by the group identif- ier. For example, Host A might employ a particular group-id, for which it maintains a list of explicit users in that group, in its network exchange with Host B, which accepts the group-id under the conditions of this interpretation. For networks, individual hosts will impose need-to-know controls over their users on the basis of named individuals - much like (in fact, probably the same) controls used when there is no network connection. When group identifiers are acceptable for access con- trol, the identifier of some other host may be employed, to eliminate the maintenance that would be required if indivi- dual identification of remote users was employed. In class C2 and higher, however, it must be possible from that audit record to identify (immediately or at some later time) exactly the individuals represented by a group identifier at the time of the use of that identifier. There is allowed to be an uncertainty because of elapsed time between changes in the group membership and the enforcement in the access con- trol mechanisms. The DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. The reference monitor manages all the physical resources of the system and from them creates the abstraction of subjects and objects that it con- trols. Some of these subjects and objects may be used to implement a part of the NTCB. When the DAC mechanism is distributed in such NTCB subjects (i.e., when outside the reference monitor), the assurance requirements (see the Assurance section) for the design and implementation of the DAC shall be those of class C2 for all networks of class C2 or above. When integrity is included as part of the network dis- cretionary security policy, the above interpretations shall be specifically applied to the controls over modification, viz, the write mode of access, within each component based on identified users or groups of users. + Rationale In this class, the supporting elements of the overall DAC mechanism are required to isolate information (objects) that supports DAC so that it is subject to auditing require- ments (see the System Architecture section). The use of network identifiers to identify groups of individual users could be implemented, for example, as an X.25 community of interest in the network protocol layer (layer 3). In all other respects, the supporting elements of the overall DAC mechanism are treated exactly as untrusted subjects are treated with respect to DAC in an ADP system, with the same result as noted in the interpretation. A typical situation for DAC is that a surrogate process for a remote user will be created in some host for access to objects under the control of the NTCB partition within that host. The interpretation requires that a user identifier be assigned and maintained for each such process by the NTCB, so that access by a surrogate process is subject to essen- tially the same discretionary controls as access by a pro- cess acting on behalf of a local user would be. However, within this interpretation a range of possible interpreta- tions of the assigned user identification is permitted. The most obvious situation would exist if a global database of network users were to be made permanently avail- able on demand to every host, (i.e., a name server existed) so that all user identifications were globally meaningful. It is also acceptable, however, for some NTCB parti- tions to maintain a database of locally-registered users for its own use. In such a case, one could choose to inhibit the creation of surrogate processes for locally unregistered users, or (if permitted by the local policy) alternatively, to permit the creation of surrogate processes with preselected user and group identifiers which, in effect, identify the process as executing on behalf of a member of a group of users on a particular remote host. The intent of the words concerning audit in the interpretation is to pro- vide a minimally acceptable degree of auditability for cases such as the last described. What is required is that there be a capability, using the audit facilities provided by the network NTCB partitions involved, to determine who was logged in at the actual host of the group of remote users at the time the surrogate processing occured. Associating the proper user id with a surrogate process is the job of identification and authentication. This means that DAC is applied locally, with respect to the user id of the surrogate process. The transmission of the data back across the network to the user's host, and the creation of a copy of the data there, is not the business of DAC. Components that support only internal subjects impact the implementation of the DAC by providing services by which information (e.g., a user-id) is made available to a com- ponent that makes a DAC decision. An example of the latter would be the case that a user at Host A attempts to access a file at Host B. The DAC decision might be (and usually would be) made at Host B on the basis of a user-id transmit- ted from Host A to Host B. Unique user identification may be achieved by a variety of mechanisms, including (a) a requirement for unique iden- tification and authentication on the host where access takes place; (b) recognition of fully qualified network addresses authenticated by another host and forwarded to the host where access takes place; or (c) administrative support of a network-wide unique personnel identifier that could be authenticated and forwarded by another host as in (b) above, or could be authenticated and forwarded by a dedicated net- work identification and authentication server. The proto- cols which implement (b) or (c) are subject to the System Architecture requirements. Network support for DAC might be handled in other ways than that described as "typical" above. In particular, some form of centralized access control is often proposed. An access control center may make all decisions for DAC, or it may share the burden with the hosts by controlling host-to- host connections, and leaving the hosts to decide on access to their objects by users at a limited set of remote hosts. In this case the access control center provides the linkage between the connection oriented abstraction (as discussed in the Introduction) and the overall network security policy for DAC. In all cases the enforcement of the decision must be provided by the host where the object resides. There are two forms of distribution for the DAC mechan- ism: implementing portions of the DAC in separate com- ponents, and supporting the DAC in subjects contained within the NTCB partition in a component. Since "the ADP system" is understood to be "the computer network" as a whole, each network component is responsible for enforcing security in the mechanisms allocated to it to ensure secure implementa- tion of the network security policy. For traditional host systems it is frequently easy to also enforce the DAC along with the MAC within the reference monitor, per se, although a few approaches, such as virtual machine monitors, support DAC outside this interface. In contrast to the universally rigid structure of man- datory policies (see the Mandatory Access Control section), DAC policies tend to be very network and system specific, with features that reflect the natural use of the system. For networks it is common that individual hosts will impose controls over their local users on the basis of named individuals-much like the controls used when there is no network connection. However, it is difficult to manage in a centralized manner all the individuals using a large net- work. Therefore, users on other hosts are commonly grouped together so that the controls required by the network DAC policy are actually based on the identity of the hosts or other components. A gateway is an example of such a com- ponent. The assurance requirements are at the very heart of the concept of a trusted system. It is the assurance that determines if a system or network is appropriate for a given environment, as reflected, for example, in the Environments Guideline-. In the case of monolithic systems that have DAC integral to the reference monitor, the assurance require- ments for DAC are inseparable from those of the rest of the reference monitor. For networks there is typically a much clearer distinction due to distributed DAC. The rationale for making the distinction in this network interpretation is that if major trusted network components can be made signi- ficantly easier to design and implement without reducing the ability to meet security policy, then trusted networks will be more easily available. 3.3.1.2 Object Reuse + Statement from DoD 5200.28-STD All authorizations to the information contained within a storage object shall be revoked prior to initial assignment, allocation or reallocation to a subject from the TCB's pool of unused storage objects. No information, including encrypted representations of information, produced by a prior subject's actions is to be available to any subject that obtains access to an object that has been released back to the system. + Interpretation The NTCB shall ensure that any storage objects that it controls (e.g., message buffers under the control of a NTCB _________________________ - Guidance for Applying the Department of Defense ________ ___ ________ ___ __________ __ _______ Trusted Computer System Evaluation Criteria in Specific _______ ________ ______ __________ ________ __ ________ Environments, CSC-STD-003-85. ____________ partition in a component) contain no information for which a subject in that component is not authorized before granting access. This requirement must be enforced by each of the NTCB partitions. + Rationale In a network system, storage objects of interest are things that the NTCB directly controls, such as message buffers in components. Each component of the network system must enforce the object reuse requirement with respect to the storage objects of interest as determined by the network security policy. For example, the DAC requirement in this division leads to the requirement here that message buffers be under the control of the NTCB partition. A buffer assigned to an internal subject may be reused at the discre- tion of that subject which is responsible for preserving the integrity of message streams. Such controlled objects may be implemented in physical resources, such as buffers, disk sectors, tape space, and main memory, in components such as network switches. 3.3.1.3 Labels + Statement from DoD 5200.28-STD Sensitivity labels associated with each ADP system resource (e.g., subject, storage object, ROM) that is directly or indirectly accessible by subjects external to the TCB shall be maintained by the TCB. These labels shall be used as the basis for mandatory access control decisions. In order to import non-labeled data, the TCB shall request and receive from an authorized user the sensitivity level of the data, and all such actions shall be auditable by the TCB. + Interpretation Non-labeled data imported under the control of the NTCB partition will be assigned a label constrained by the device labels of the single-level device used to import it. Labels may include secrecy and integrity- components in accordance with the overall network security policy described by the network sponsor. Whenever the term "label" is used throughout this interpretation, it is understood to include both components as applicable. Similarly, the terms "single-level" and "multilevel" are understood to be based on both the secrecy and integrity components of the policy. The mandatory integrity policy will typically have require- ments, such as the probability of undetected message stream modification, that will be reflected in the label for the _________________________ - See, for example, Biba, K.J., "Integrity Considera- tion for Secure Computer Systems," ESD-TR-76-372, MTR- 3153, The MITRE Corporation, Bedford, MA, April 1977. data so protected. For example, when data is imported its integrity label may be assigned based on mechanisms, such as cryptography, used to provide the assurance required by the policy. The NTCB shall assure that such mechanism are pro- tected from tampering and are always invoked when they are the basis for a label. If the security policy includes an integrity policy, all activities that result in message-stream modification during transmission are regarded as unauthorized accesses in violation of the integrity policy. The NTCB shall have an automated capability for testing, detecting, and reporting those errors/corruptions that exceed specified network integrity policy requirements. Message-stream modification (MSM) countermeasures shall be identified. A technology of adequate strength shall be selected to resist MSM. If encryption methodologies are employed, they shall be approved by the National Security Agency. All objects must be labeled within each component of the network that is trusted to maintain separation of multi- ple levels of information. The label associated with any objects associated with single-level components will be identical to the level of that component. Objects used to store network control information, and other network struc- tures, such as routing tables, must be labeled to prevent unauthorized access and/or modification. + Rationale The interpretation is an extension of the requirement into the context of a network system and partitioned NTCB as defined for these network interpretations. A single-level device may be regarded either as a subject or an object. A multilevel device is regarded as a trusted subject in which the security range of the subject is the minimum-maximum range of the data expected to be transmitted over the dev- ice. The sensitivity labels for either secrecy or integrity or both may reflect non-hierarchical categories or hierarch- ical classification or both. For a network it is necessary that this requirement be applied to all network system resources at the (B2) level and above. The NTCB is responsible for implementing the network integrity policy, when one exists. The NTCB must enforce that policy by ensuring that information is accurately transmitted from source to destination (regardless of the number of intervening connecting points). The NTCB must be able to counter equipment failure, environmental disrup- tions, and actions by persons and processes not authorized to alter the data. Protocols that perform code or format conversion shall preserve the integrity of data and control information. The probability of an undetected transmission error may be specified as part of the network security policy so that the acceptability of the network for its intended applica- tion may be determined. The specific metrics (e.g., proba- bility of undetected modification) satisfied by the data can be reflected in the integrity sensitivity label associated with the data while it is processed within a component. It is recognized that different applications and operational environments (e.g., crisis as compared to logistic) will have different integrity requirements. The network shall also have an automated capability of testing for, detecting, and reporting errors that exceed a threshold consistent with the operational mode requirements. The effectiveness of integrity countermeasures must be esta- blished with the same rigor as the other security-relevant properties such as secrecy. Cryptography is often utilized as a basis to provide data integrity assurance. Mechanisms, such as Manipulation Detection Codes (MDC)-, may be used. The adequacy of the encryption or MDC algorithm, the correctness of the protocol logic, and the adequacy of implementation must be esta- blished in MSM countermeasures design. 3.3.1.3.1 Label Integrity + Statement from DoD 5200.28-STD Sensitivity labels shall accurately represent sensitivity levels of the specific subjects or objects with which they are associated. When exported by the TCB, sensitivity labels shall accurately and unambiguously represent the internal labels and shall be associated with the information being exported. + Interpretation The phrase "exported by the TCB" is understood to include transmission of information from an object in one component to an object in another component. Information transferred between NTCB partitions is addressed in the Sys- tem Integrity Section. The form of internal and external (exported) sensitivity labels may differ, but the meaning shall be the same. The NTCB shall, in addition, ensure that correct association of sensitivity labels with the informa- tion being transported across the network is preserved. _________________________ - See Jueneman, R. R., "Electronic Document Authenti- cation," IEEE Network Magazine, April 1987, pp 17-23. ____ _______ ________ As mentioned in the Trusted Facility Manual Section, encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity level of the ciphertext is generally lower than the cleartext. It fol- lows that cleartext and ciphertext are contained in dif- ferent objects, each possessing its own label. The label of the cleartext must be preserved and associated with the ciphertext so that it can be restored when the cleartext is subsequently obtained by decrypting the ciphertext. If the cleartext is associated with a single-level device, the label of that cleartext may be implicit. The label may also be implicit in the key. When information is exported to an environment where it is subject to deliberate or accidental modification, the TCB shall support the means, such as cryptographic checksums, to assure the accuracy of the labels. When there is a manda- tory integrity policy, the policy will define the meaning of integrity labels. + Rationale Encryption algorithms and their implementation are out- side the scope of these interpretations. Such algorithms may be implemented in a separate device or may be incor- porated in a subject of a larger component. Without preju- dice, either implementation packaging is referred to as an encryption mechanism herein. If encryption methodologies are employed in this regard, they shall be approved by the National Security Agency (NSA). The encryption process is part of the Network Trusted Computer Base partition in the components in which it is implemented. The encryption mechanism is not necessarily a mul- tilevel device or multilevel subject, as these terms are used in these criteria. The process of encryption is mul- tilevel by definition. The cleartext and ciphertext inter- faces carry information of different sensitivity. An encryption mechanism does not process data in the sense of performing logical or arithmetic operations on that data with the intent of producing new data. The cleartext and ciphertext interfaces on the encryption mechanism must be separately identified as being single-level or multilevel. If the interface is single-level, then the sensitivity of the data is established by a trusted individual and impli- citly associated with the interface; the Exportation to Single-Level Devices criterion applies. If the interface is multilevel, then the data must be labeled; the Exportation to Multilevel Devices criterion applies. The network architect is free to select an accept- able mechanism for associating a label with an object. With reference to encrypted objects, the following examples are possible: 1. Include a label field in the protocol definition of the object. 2. Implicitly associate the label with the object through the encryption key. That is, the encryption key uniquely identifies a sensitivity level. A sin- gle or private key must be protected at the level of the data that it encrypts. 3.3.1.3.2 Exportation of Labeled Information + Statement from DoD 5200.28-STD The TCB shall designate each communication channel and I/O device as either single-level or multilevel. Any change in this designation shall be done manually and shall be audit- able by the TCB. The TCB shall maintain and be able to audit any change in the sensitivity level or levels associ- ated with a communications channel or I/O device. + Interpretation Each communication channel and network component shall be designated as either single-level or multilevel. Any change in this designation shall be done with the cognizance and approval of the administrator or security officer in charge of the affected components and the administrator or security officer in charge of the NTCB. This change shall be auditable by the network. The NTCB shall maintain and be able to audit any change in the device labels associated with a single-level communication channel or the range asso- ciated with a multilevel communication channel or component. The NTCB shall also be able to audit any change in the set of sensitivity levels associated with the information which can be transmitted over a multilevel communication channel or component. + Rationale Communication channels and components in a network are analogous to communication channels and I/O devices in stand-alone systems. They must be designated as either mul- tilevel (i.e., able to distinguish and maintain separation among information of various sensitivity levels) or single- level. As in the TCSEC, single-level devices may only be attached to single-level channels. The level or set of levels of information that can be sent to a component or over a communication channel shall only change with the knowledge and approval of the security officers (or system administrator, if there is no security officer) of the network, and of the affected components. This requirement ensures that no significant security- relevant changes are made without the approval of all affected parties. 3.3.1.3.2.1 Exportation to Multilevel Devices + Statement from DoD 5200.28-STD When the TCB exports an object to a multilevel I/O device, the sensitivity label associated with that object shall also be exported and shall reside on the same physical medium as the exported information and shall be in the same form (i.e., machine-readable or human-readable form). When the TCB exports or imports an object over a multilevel communi- cations channel, the protocol used on that channel shall provide for the unambiguous pairing between the sensitivity labels and the associated information that is sent or received. + Interpretation The components, including hosts, of a network shall be interconnected over "multilevel communication channels," multiple single-level communication channels, or both, when- ever the information is to be protected at more than a sin- gle sensitivity level. The protocol for associating the sensitivity label and the exported information shall provide the only information needed to correctly associate a sensi- tivity level with the exported information transferred over the multilevel channel between the NTCB partitions in indi- vidual components. This protocol definition must specify the representation and semantics of the sensitivity labels (i.e., the machine-readable label must uniquely represent the sensitivity level). The "unambiguous" association of the sensitivity level with the communicated information shall meet the same level of accuracy as that required for any other label within the NTCB, as specified in the criterion for Label Integrity. This may be provided by protected and highly reliable direct physical layer connections, or by traditional cryptographic link protection in which any errors during transmission can be readily detected, or by use of a separate channel. The range of information imported or exported must be con- strained by the associated device labels. + Rationale This protocol must specify the representation and semantics of the sensitivity labels. See the Mandatory Access Control Policies section in Appendix B. The mul- tilevel device interface to (untrusted) subjects may be implemented either by the interface of the reference moni- tor, per se, or by a multilevel subject (e.g., a "trusted subject" as defined in the Bell-LaPadula Model) that pro- vides the labels based on the internal labels of the NTCB partition. The current state of the art limits the support for mandatory policy that is practical for secure networks. Reference monitor support to ensure the control over all the operations of each subject in the network must be completely provided within the single NTCB partition on which that sub- ject interfaces to the NTCB. This means that the entire portion of the "secure state" represented in the formal security policy model that may be changed by transitions invoked by this subject must be contained in the same com- ponent. The secure state of an NTCB partition may be affected by events external to the component in which the NTCB parti- tion resides (e.g., arrival of a message). The effect occurs asynchronusly after being initiated by an event in another component or partition. For example, indeterminate delays may occur between the initiation of a message in one component, the arrival of the message in the NTCB partition in another component, and the corresponding change to the secure state of the second component. Since each component is executing concurrently, to do otherwise would require some sort of network-wide control to synchronize state tran- sitions, such as a global network-wide clock for all proces- sors; in general, such designs are not practical and prob- ably not even desirable. Therefore, the interaction between NTCB partitions is restricted to just communications between pairs (at least logically) of devices-multilevel devices if the device(s) can send/receive data of more than a single level. For broadcast channels the pairs are the sender and intended receiver(s). However, if the broadcast channel carries multiple levels of information, additional mechanism (e.g., cryptochecksum maintained by the TCB) may be required to enforce separation and proper delivery. A common representation for sensitivity labels is needed in the protocol used on that channel and understood by both the sender and receiver when two multilevel devices (in this case, in two different components) are intercon- nected. Each distinct sensitivity level of the overall net- work policy must be represented uniquely in these labels. Within a monolithic TCB, the accuracy of the sensi- tivity labels is generally assured by simple techniques, e.g., very reliable connections over very short physical connections, such as on a single printed circuit board or over an internal bus. In many network environments there is a much higher probability of accidentally or maliciously introduced errors, and these must be protected against. 3.3.1.3.2.2 Exportation to Single-Level Devices + Statement from DoD 5200.28-STD Single-level I/O devices and single-level communication channels are not required to maintain the sensitivity labels of the information they process. However, the TCB shall include a mechanism by which the TCB and an authorized user reliably communicate to designate the single sensitivity level of information imported or exported via single-level communication channels or I/O devices. + Interpretation Whenever one or both of two directly connected com- ponents is not trusted to maintain the separation of infor- mation of different sensitivity levels, or whenever the two directly connected components have only a single sensitivity level in common, the two components of the network shall communicate over a single-level channel. Single-level com- ponents and single-level communication channels are not required to maintain the sensitivity labels of the informa- tion they process. However, the NTCB shall include a reli- able communication mechanism by which the NTCB and an authorized user (via a trusted path) or a subject within an NTCB partition can designate the single sensitivity level of information imported or exported via single-level communica- tion channels or network components. The level of informa- tion communicated must equal the device level. + Rationale Single-level communications channels and single-level components in networks are analogous to single level chan- nels and I/O devices in stand-alone systems in that they are not trusted to maintain the separation of information of different sensitivity levels. The labels associated with data transmitted over those channels and by those components are therefore implicit; the NTCB associates labels with the data because of the channel or component, not because of an explicit part of the bit stream. Note that the sensitivity level of encrypted information is the level of the cipher- text rather than the original level(s) of the plaintext. 3.3.1.3.2.3 Labeling Human-Readable Output + Statement from DoD 5200.28-STD The ADP system administrator shall be able to specify the printable label names associated with exported sensitivity labels. The TCB shall mark the beginning and end of all human-readable, paged, hardcopy output (e.g., line printer output) with human-readable sensitivity labels that prop- erly1 represent the sensitivity of the output. The TCB shall, by default, mark the top and bottom of each page of human-readable, paged, hardcopy output (e.g., line printer output) with human-readable sensitivity labels that prop- erly1 represent the sensitivity of the page. The TCB shall, by default and in an appropriate manner, mark other forms of human readable output (e.g., maps, graphics) with human- readable sensitivity labels that properly1 represent the sensitivity of the output. Any override of these markings defaults shall be auditable by the TCB. _________________________ 1 The hierarchical classification component in human- readable sensitivity labels shall be equal to the greatest hierarchical classification of any of the information in the output that the labels refer to; the non-hierarchical category component shall include all of the non-hierarchical categories of the information in the output the labels refer to, but no other non-hierarchical categories. + Interpretation This criterion imposes no requirement to a component that produces no human-readable output. For those that do produce human-readable output, each sensitivity level that is defined to the network shall have a uniform meaning across all components. The network administrator, in con- junction with any affected component administrator, shall be able to specify the human-readable label that is associated with each defined sensitivity level. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system and partitioned NTCB as defined for these network interpreta- tions. 3.3.1.3.3 Subject Sensitivity Labels + Statement from DoD 5200.28-STD The TCB shall immediately notify a terminal user of each change in the sensitivity level associated with that user during an interactive session. A terminal user shall be able to query the TCB as desired for a display of the subject's complete sensitivity label. + Interpretation An NTCB partition shall immediately notify a terminal user attached to its component of each change in the sensi- tivity level associated with that user. + Rationale The local NTCB partition must ensure that the user understands the sensitivity level of information sent to and from a terminal. When a user has a surrogate process in another component, adjustments to its level may occur to maintain communication with the user. These changes may occur asynchronously. Such adjustments are necessitated by mandatory access control as applied to the objects involved in the communication path. 3.3.1.3.4 Device Labels + Statement from DoD 5200.28-STD The TCB shall support the assignment of minimum and maximum sensitivity levels to all attached physical devices. These sensitivity levels shall be used by the TCB to enforce con- straints imposed by the physical environments in which the devices are located. + Interpretation This requirement applies as written to each NTCB parti- tion that is trusted to separate information based on sensi- tivity level. Each I/O device in a component, used for com- munication with other network components, is assigned a dev- ice range, consisting of a set of labels with a maximum and minimum. (A device range usually contains, but does not necessarily contain, all possible labels "between" the max- imum and minimum, in the sense of dominating the minimum and being dominated by the maximum.) The NTCB always provides an accurate label for informa- tion exported through devices. Information exported or imported using a single-level device is labelled implicitly by the sensitivity level of the device. Information exported from one multilevel device and imported at another must be labelled through an agreed-upon protocol, unless it is labelled implicitly by using a communication link that always carries a single level. Information exported at a given sensitivity level can be sent only to an importing device whose device range con- tains that level or a higher level. If the importing device range does not contain the given level, the information is relabelled upon reception at a higher level within the importing device range. Relabelling should not occur other- wise. + Rationale The purpose of device labels is to reflect and con- strain the sensitivity levels of information authorized for the physical environment in which the devices are located. The information transfer restrictions permit one-way communication (i.e., no acknowledgements) from one device to another whose ranges have no level in common, as long as each level in the sending device range is dominated by some level in the receiving device range. It is never permitted to send information at a given level to a device whose range does not contain a dominating level. (See Appendix C for similar interconnection rules for the interconnected AIS view.) 3.3.1.4 Mandatory Access Control + Statement from DoD 5200.28-STD The TCB shall enforce a mandatory access control policy over all resources (i.e., subjects, storage objects, and I/O dev- ices) that are directly or indirectly accessible by subjects external to the TCB. These subjects and objects shall be assigned sensitivity labels that are a combination of hierarchical classification levels and non-hierarchical categories, and the labels shall be used as the basis for mandatory access control decisions. The TCB shall be able to support two or more such sensitivity levels. (See the Mandatory Access Control interpretations.) The following requirements shall hold for all accesses between all sub- jects external to the TCB and all objects directly or indirectly accessible by these subjects. A subject can read an object only if the hierarchical classification in the subject's sensitivity level is greater than or equal to the hierarchical classification of the object's sensitivity level and the non-hierarchical categories in the subject's sensitivity level include all the non-hierarchical categories in the object's sensitivity level. A subject can write an object only if the hierarchical classification in the subject's sensitivity level is less than or equal to the hierarchical classification of the object's sensitivity level and the non-hierarchical categories in the subject's sensitivity level are included in the non-hierarchical categories in the object's sensitivity level. Identification and authentication data shall be used by the TCB to authen- ticate the user's identity and to ensure that the sensi- tivity level and authorization of subjects external to the TCB that may be created to act on behalf of the individual user are dominated by the clearance and authorization of that user. + Interpretation Each partition of the NTCB exercises mandatory access control policy over all subjects and objects in its com- ponent. In a network, the responsibility of an NTCB parti- tion encompasses all mandatory access control functions in its component that would be required of a TCB in a stand- alone system. In particular, subjects and objects used for communication with other components are under the control of the NTCB partition. Mandatory access control includes secrecy and integrity control to the extent that the network sponsor has described in the overall network security pol- icy. Conceptual entities associated with communication between two components, such as sessions, connections and virtual circuits, may be thought of as having two ends, one in each component, where each end is represented by a local object. Communication is viewed as an operation that copies information from an object at one end of a communication path to an object at the other end. Transient data-carrying entities, such as datagrams and packets, exist either as information within other objects, or as a pair of objects, one at each end of the communication path. The requirement for "two or more" sensitivity levels can be met by either secrecy or integrity levels. When there is a mandatory integrity policy, the stated require- ments for reading and writing are generalized to: A subject can read an object only if the subject's sensitivity level dominates the object's sensitivity level, and a subject can write an object only if the object's sensitivity level dom- inates the subject's sensitivity level. Based on the integrity policy, the network sponsor shall define the domi- nance relation for the total label, for example, by combin- ing secrecy and integrity lattices. - + Rationale An NTCB partition can maintain access control only over subjects and objects in its component. At levels B2 and above, the NTCB partition must maintain access control over all subjects and objects in its component. Access by a sub- ject in one component to information contained in an object in another component requires the creation of a subject in the remote component which acts as a surrogate for the first subject. The mandatory access controls must be enforced at the interface of the reference monitor (viz. the mechanism that controls physical processing resources) for each NTCB parti- tion. This mechanism creates the abstraction of subjects and objects which it controls. Some of these subjects out- side the reference monitor, per se, may be designated to implement part of an NTCB partition's mandatory policy, e.g., by using the ``trusted subjects" defined in the Bell- LaPadula model. The prior requirements on exportation of labeled infor- mation to and from I/O devices ensure the consistency between the sensitivity labels of objects connected by a communication path. As noted in the introduction, the net- work architecture must recognize the linkage between the overall mandatory network security policy and the connection oriented abstraction. For example, individual data-carrying entities such as datagrams can have individual sensitivity labels that subject them to mandatory access control in each component. The abstraction of a single-level connection is realized and enforced implicitly by an architecture while a connection is realized by single-level subjects that neces- sarily employ only datagrams of the same level. The fundamental trusted systems technology permits the DAC mechanism to be distributed, in contrast to the require- ments for mandatory access control. For networks this separation of MAC and DAC mechanisms is the rule rather than _________________________ - See, for example, Grohn, M. J., A Model of a Pro- _ _____ __ _ ___ tected Data Management System, ESD-TR-76-289, I. P. ______ ____ __________ ______ Sharp Assoc. Ltd., June, 1976; and Denning, D .E., Lunt, T. F., Neumann, P. G., Schell, R. R., Heckman, M. and Shockley, W., Secure Distributed Data Views, Secu- ______ ___________ ____ _____ ____ rity Policy and Interpretation for a Class A1 Multilev- ____ ______ ___ ______________ ___ _ _____ __ ________ el Secure Relational Database System,SRI International, __ ______ __________ ________ ______ November 1986. the exception. The set of total sensitivity labels used to represent all the sensitivity levels for the mandatory access control (combined data secrecy and data integrity) policy always forms a partially ordered set. Without loss of generality, this set of labels can always be extended to form a lattice, by including all the combinations of non-hierarchical categories. As for any lattice, a dominance relation is always defined for the total sensitivity labels. For admin- istrative reasons it may be helpful to have a maximum level which dominates all others. 3.3.2 Accountability _ _ _ ______________ 3.3.2.1 Identification and Authentication + Statement from DoD 5200.28-STD The TCB shall require users to identify themselves to it before beginning to perform any other actions that the TCB is expected to mediate. Furthermore, the TCB shall maintain authentication data that includes information for verifying the identify of individual users (e.g., passwords) as well as information for determining the clearance and authoriza- tions of individual users. This data shall be used by the TCB to authenticate the user's identity and to ensure that the sensitivity level and authorization of subjects external to the TCB that may be created to act on behalf of the indi- vidual user are dominated by the clearance and authorization of that user. The TCB shall protect authentication data so that it cannot be accessed by any unauthorized user. The TCB shall be able to enforce individual accountability by providing the capability to uniquely identify each indivi- dual ADP system user. The TCB shall also provide the capa- bility of associating this identity with all auditable actions taken by that individual. + Interpretation The requirement for identification and authentication of users is the same for a network system as for an ADP sys- tem. The identification and authentication may be done by the component to which the user is directly connected or some other component, such as an identification and authen- tication server. Available techniques, such as those described in the Password Guideline=, are generally also applicable in the network context. However, in cases where the NTCB is expected to mediate actions of a host (or other network component) that is acting on behalf of a user or group of users, the NTCB may employ identification and _________________________ = Department of Defense Password Management Guide- __________ __ _______ ________ __________ _____ line, CSC-STD-002-85 ____ authentication of the host (or other component) in lieu of identification and authentication of an individual user, so long as the component identifier implies a list of specific users uniquely associated with the identifier at the time of its use for authentication. This requirement does not apply to internal subjects. Authentication information, including the identity of a user (once authenticated) may be passed from one component to another without reauthentication, so long as the NTCB protects (e.g., by encryption) the information from unau- thorized disclosure and modification. This protection shall provide at least a similar level of assurance (or strength of mechanism) as pertains to the protection of the authenti- cation mechanism and authentication data. + Rationale The need for accountability is not changed in the con- text of a network system. The fact that the NTCB is parti- tioned over a set of components neither reduces the need nor imposes new requirements. That is, individual accountabil- ity is still the objective. Also, in the context of a net- work system at the (C2) level or higher "individual accoun- tability" can be satisfied by identification of a host (or other component) so long as the requirement for traceability to individual users or a set of specific individual users with active subjects is satisfied. There is allowed to be an uncertainty in traceability because of elapsed time between changes in the group membership and the enforcement in the access control mechanisms. In addition, there is no need in a distributed processing system like a network to reauthen- ticate a user at each point in the network where a projec- tion of a user (via the subject operating on behalf of the user) into another remote subject takes place. The passing of identifiers and/or authentication infor- mation from one component to another is usually done in sup- port to the implementation of the discretionary access con- trol (DAC). This support relates directly to the DAC regarding access by a user to a storage object in a dif- ferent NTCB partition than the one where the user was authenticated. Employing a forwarded identification implies additional reliance on the source and components along the path. If the authenticated identification is used as the basis of determining a sensitivity label for a subject, it must satisfy the Label Integrity criterion. An authenticated identification may be forwarded between components and employed in some component to iden- tify the sensitivity level associated with a subject created to act on behalf of the user so identified. 3.3.2.1.1 Trusted Path + Statement from DoD 5200.28-STD The TCB shall support a trusted communication path between itself and USERS FOR USE WHEN A POSITIVE TCB-TO-USER CONNEC- TION IS REQUIRED (E.G., LOGIN, CHANGE SUBJECT SENSITIVITY LEVEL). Communications via this TRUSTED path shall be ACTIVATED exclusively by a USER OR THE TCB AND SHALL BE LOGICALLY AND UNMISTAKABLY DISTINGUISHABLE FROM OTHER PATHS. + Interpretation A trusted path is supported between a user (i.e., human) and the NTCB partition in the component to which the user is directly connected. + Rationale When a user logs into a remote component, the user id is transmitted securely between the local and remote NTCB partitions in accordance with the requirements in Identifi- cation and Authentication. Trusted Path is necessary in order to assure that the user is communicating with the NTCB and only the NTCB when security relevant activities are taking place (e.g., authen- ticate user, set current session sensitivity level). How- ever, Trusted Path does not address communications within the NTCB, only communications between the user and the NTCB. If, therefore, a component does not support any direct user communication then the component need not contain mechanisms for assuring direct NTCB to user communications. The requirement for trusted communication between one NTCB partition and another NCTB partition is addressed in the System Architecture section. These requirements are separate and distinct from the user to NTCB communication requirement of a trusted path. However, it is expected that this trusted communication between one NTCB partition and another NTCB partition will be used in conjunction with the trusted path to implement trusted communication between the user and the remote NTCB partition. 3.3.2.2 Audit + Statement from DoD 5200.28-STD The TCB shall be able to create, maintain, and protect from modification or unauthorized access or destruction an audit trail of accesses to the objects it protects. The audit data shall be protected by the TCB so that read access to it is limited to those who are authorized for audit data. The TCB shall be able to record the following types of events: use of identification and authentication mechanisms, introduction of objects into a user's address space (e.g., file open, program initiation), deletion of objects, actions taken by computer operators and system administrators and/or system security officers, and other security relevant events. The TCB shall also be able to audit any override of human-readable output markings. For each recorded event, the audit record shall identify: date and time of the event, user, type of event, and success or failure of the event. For identification/authentication events the origin of request (e.g., terminal ID) shall be included in the audit record. For events that introduce an object into a user's address space and for object deletion events the audit record shall include the name of the object and the object's sensitivity level. The ADP system administrator shall be able to selectively audit the actions of any one or more users based on individual identify and/or object sensitivity level. The TCB shall be able to audit the identified events that may be used in the exploitation of covert storage channels. THE TCB SHALL CONTAIN A MECHANISM THAT IS ABLE TO MONITOR THE OCCURRENCE OR ACCUMULATION OF SECU- RITY AUDITABLE EVENTS THAT MAY INDICATE AN IMMINENT VIOLA- TION OF SECURITY POLICY. THIS MECHANISM SHALL BE ABLE TO IMMEDIATELY NOTIFY THE SECURITY ADMINISTRATOR WHEN THRES- HOLDS ARE EXCEEDED AND, IF THE OCCURRENCE OR ACCUMULATION OF THESE SECURITY RELEVANT EVENTS CONTINUES, THE SYSTEM SHALL TAKE THE LEAST DISRUPTIVE ACTION TO TERMINATE THE EVENT. + Interpretation This criterion applies as stated. The sponsor must select which events are auditable. If any such events are not distinguishable by the NTCB alone (for example those identified in Part II), the audit mechanism shall provide an interface, which an authorized subject can invoke with parameters sufficient to produce an audit record. These audit records shall be distinguishable from those provided by the NTCB. In the context of a network system, "other security relevant events" (depending on network system architecture and network security policy) might be as fol- lows: 1. Identification of each access event (e.g., estab- lishing a connection or a connectionless association between processes in two hosts of the network) and its principal parameters (e.g., host identifiers of the two hosts involved in the access event and user identifier or host identifier of the user or host that is requesting the access event) 2. Identification of the starting and ending times of each access event using local time or global syn- chronized time 3. Identification of security-relevant exceptional con- ditions (e.g., potential violation of data integrity, such as misrouted datagrams) detected during the transactions between two hosts 4. Utilization of cryptographic variables 5. Changing the configuration of the network (e.g., a component leaving the network and rejoining) In addition, identification information should be included in appropriate audit trail records, as necessary, to allow association of all related (e.g., involving the same network event) audit trail records (e.g., at different hosts) with each other. Furthermore, a component of the network system may provide the required audit capability (e.g., storage, retrieval, reduction, analysis) for other components that do not internally store audit data but transmit the audit data to some designated collection com- ponent. Provisions shall be made to control the loss of audit data due to unavailability of resources. In the context of a network system, the "user's address space" is extended, for object introduction and deletion events, to include address spaces being employed on behalf of a remote user (or host). However, the focus remains on users in contrast to internal subjects as discussed in the DAC criterion. In addition, audit information must be stored in machine-readable form. The capability must exist to audit the identified events that may be used in the exploitation of covert storage channels. To accomplish this, each NTCB partition must be able to audit those events locally that may lead to the exploitation of a covert storage channel which exist because of the network. THE SPONSOR SHALL IDENTIFY THE SPECIFIC AUDITABLE EVENTS THAT MAY INDICATE AN IMMINENT VIOLATION OF SECURITY POLICY. THE COMPONENT WHICH DETECTS THE OCCURRENCE OR ACCU- MULATION OF SUCH EVENTS MUST BE ABLE TO NOTIFY AN APPROPRI- ATE ADMINISTRATOR WHEN THRESHOLDS ARE EXCEEDED, AND TO INI- TIATE ACTIONS WHICH WILL RESULT IN TERMINATION OF THE EVENT IF THE ACCUMULATION CONTINUES. FOR EXAMPLE, WHEN THE THRES- HOLD OF UNSUCCESSFUL LOGIN ATTEMPTS WITHIN A PERIOD OF TIME IS EXCEEDED, LOGIN SHALL BE INHIBITED FOR A SPECIFIC TIME PERIOD. + Rationale For remote users, the network identifiers (e.g., inter- net address) can be used as identifiers of groups of indivi- dual users (e.g., "all users at Host A") to eliminate the maintenance that would be required if individual identifica- tion of remote users was employed. In this class (C2), how- ever, it must be possible to identify (immediately or at some later time) the individuals represented by a group identifier. In all other respects, the interpretation is a straightforward extension of the criterion into the context of a network system. Identification of covert channel events is addressed in the Covert Channel Analysis section. BECAUSE OF CONCURRENCY AND SYNCHRONIZATION PROBLEMS, IT MAY NOT BE POSSIBLE TO DETECT IN REAL TIME THE ACCUMULATION OF SECURITY AUDITABLE EVENTS THAT ARE OCCURRING IN DIFFERENT NTCB PARTITIONS. HOWEVER, EACH NTCB PARTITION THAT HAS BEEN ALLOCATED AUDIT RESPONSIBILITY MUST HAVE THE CAPABILITY TO DETECT THE LOCAL ACCUMULATION OF EVENTS, TO NOTIFY THE PAR- TITION SECURITY ADMINISTRATOR AND/OR THE NETWORK SECURITY ADMINISTRATOR, AND TO INITIATE ACTIONS WHICH WILL RESULT IN TERMINATION OF THE EVENT LOCALLY. 3.3.3 Assurance _ _ _ _________ 3.3.3.1 Operational Assurance 3.3.3.1.1 System Architecture + Statement from DoD 5200.28-STD The TCB shall maintain a domain for its own execution that protects it from external interference or tampering (e.g., by modification of its code or data structures). The TCB shall maintain process isolation through the provision of distinct address spaces under its control. The TCB shall be internally structured into well-defined largely independent modules. It shall make effective use of available hardware to separate those elements that are protection-critical from those that are not. The TCB modules shall be designed such that the principle of least privilege is enforced. Features in hardware, such as segmentation, shall be used to support logically distinct storage objects with separate attributes (namely: readable, writable). The user interface to the TCB shall be completely defined and all elements of the TCB identified. THE TCB SHALL BE DESIGNED AND STRUCTURED TO USE A COMPLETE, CONCEPTUALLY SIMPLE PROTECTION MECHANISM WITH PRECISELY DEFINED SEMANTICS. THIS MECHANISM SHALL PLAY A CENTRAL ROLE IN ENFORCING THE INTERNAL STRUCTURING OF THE TCB AND THE SYSTEM. THE TCB SHALL INCORPORATE SIGNIFICANT USE OF LAYERING, ABSTRACTION AND DATA HIDING. SIGNIFICANT SYSTEM ENGINEERING SHALL BE DIRECTED TOWARD MINIMIZING THE COMPLEXITY OF THE TCB AND EXCLUDING FROM THE TCB MODULES THAT ARE NOT PROTECTION-CRITICAL. + Interpretation The system architecture criterion must be met individu- ally by all NTCB partitions. Implementation of the require- ment that the NTCB maintain a domain for its own execution is achieved by having each NTCB partition maintain a domain for its own execution. Since each component is itself a dis- tinct domain in the overall network system, this also satis- fies the requirement for process isolation through distinct address spaces in the special case where a component has only a single subject. The NTCB must be internally structured into well- defined largely independent modules and meet the hardware requirements. This is satisfied by having each NTCB parti- tion so structured. The NTCB controls all network resources. These resources are the union of the sets of resources over which the NTCB partitions have control. Code and data structures belonging to the NTCB, transferred among NTCB subjects (i.e., subjects outside the reference monitor but inside the NTCB) belonging to different NTCB partitions, must be protected against external interference or tamper- ing. For example, a cryptographic checksum or physical means may be employed to protect user authentication data exchanged between NTCB partitions. Each NTCB partition must enforce the principle of least privilege within its component. Additionally, the NTCB must be structured so that the principle of least privilege is enforced in the system as a whole. THE NTCB MUST BE DESIGNED AND STRUCTURED ACCORDING TO THE NETWORK SECURITY ARCHITECTURE TO USE A COMPLETE, CONCEP- TUALLY SIMPLE PROTECTION MECHANISM. FURTHERMORE, EACH NTCB PARTITION MUST ALSO BE SO DESIGNED AND STRUCTURED. SIGNIFICANT SYSTEM ENGINEERING SHOULD BE DIRECTED TOWARD MINIMIZING THE COMPLEXITY OF EACH NTCB PARTITION, AND OF THE NTCB. CARE SHALL BE TAKEN TO EXCLUDE MODULES (AND COMPONENTS) THAT ARE NOT PROTECTION-CRITICAL FROM THE NTCB. IT IS RECOGNIZED THAT SOME MODULES AND/OR COMPONENTS MAY NEED TO BE INCLUDED IN THE NTCB AND MUST MEET THE NTCB REQUIREMENTS EVEN THOUGH THEY MAY NOT APPEAR TO BE DIRECTLY PROTECTION-CRITICAL. THE CORRECT OPERATION OF THESE MODULES/COMPONENTS IS NECESSARY FOR THE CORRECT OPERATION OF THE PROTECTION-CRITICAL MODULES AND COMPONENTS. HOWEVER, THE NUMBER AND SIZE OF THESE MODULES/COMPONENTS SHOULD BE KEPT TO A MINIMUM. Each NTCB partition provides isolation of resources (within its component) in accord with the network system architecture and security policy so that "supporting ele- ments" (e.g., DAC and user identification) for the security mechanisms of the network system are strengthened compared to C2, from an assurance point of view, through the provi- sion of distinct address spaces under control of the NTCB. As discussed in the Discretionary Access Control sec- tion, the DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. When distributed in NTCB sub- jects (i.e., when outside the reference monitor), the assurance requirements for the design and implementation of the DAC shall be those of class C2 for all networks of class C2 or above. + Rationale The requirement that the NTCB be structured into modules and meet the hardware requirements applies within the NTCB partitions in the various components. The principle of least privilege requires that each user or other individual with access to the system be given only those resources and authorizations required for the performance of this job. In order to enfore this principle in the system it must be enforced in every NTCB partition that supports users or other individuals. For example, prohibiting access by administrators to objects outside the NTCB partition (e.g., games) lessens the opportunity of dam- age by a Trojan Horse. The requirement for the protection of communications between NTCB partitions is specifically directed to subjects that are part of the NTCB partitions. Any requirements for such protection for the subjects that are outside the NTCB partitions are addressed in response to the integrity requirements of the security policy. THERE ARE CERTAIN PARTS OF A NETWORK (MODULES AND/OR COMPONENTS) THAT MAY NOT APPEAR TO BE DIRECTLY PROTECTION- CRITICAL IN THAT THEY ARE NOT INVOLVED IN ACCESS CONTROL DECISIONS, DO NOT DIRECTLY AUDIT, AND ARE NOT INVOLVED IN THE IDENTIFICATION/AUTHENTICATION PROCESS. HOWEVER, THE SECURITY OF THE NETWORK MUST DEPEND ON THE CORRECT OPERATION OF THESE MODULES AND/OR COMPONENTS. AN EXAMPLE OF THIS IS A SINGLE LEVEL PACKET SWITCH. ALTHOUGH IT MAY NOT NORMALLY BE INVOLVED DIRECTLY IN ENFORCING THE DISCRETIONARY SECURITY POLICY, THIS SWITCH MAY BE TRUSTED NOT TO MIX DATA FROM DIF- FERENT MESSAGE STREAMS. IF THE SWITCH DOES NOT OPERATE CORRECTLY, DATA COULD GET MIXED, AND UNAUTHORIZED ACCESS COULD RESULT. THEREFORE, THESE MODULES/COMPONENTS MUST BE INCLUDED IN THE NTCB AND MUST MEET THE NTCB REQUIREMENTS APPLICABLE TO THE POLICY ELEMENT(S) FOR WHICH THEY ARE RESPONSIBLE. 3.3.3.1.2 System Integrity + Statement from DoD 5200.28-STD Hardware and/or software features shall be provided that can be used to periodically validate the correct operation of the on-site hardware and firmware elements of the TCB. + Interpretation Implementation of the requirement is partly achieved by having hardware and/or software features that can be used to periodically validate the correct operation of the hardware and firmware elements of each component's NTCB partition. Features shall also be provided to validate the identity and correct operation of a component prior to its incorporation in the network system and throughout system operation. For example, a protocol could be designed that enables the com- ponents of the partitioned NTCB to exchange messages period- ically and validate each other's correct response. The pro- tocol shall be able to determine the remote entity's ability to respond. NTCB partitions shall provide the capability to report to network administrative personnel the failures detected in other NTCB partitions. Intercomponent protocols implemented within a NTCB shall be designed in such a way as to provide correct opera- tion in the case of failures of network communications or individual components. The allocation of mandatory and dis- cretionary access control policy in a network may require communication between trusted subjects that are part of the NTCB partitions in different components. This communication is normally implemented with a protocol between the subjects as peer entities. Incorrect access within a component shall not result from failure of an NTCB partition to communicate with other components. + Rationale The first paragraph of the interpretation is a straightforward extension of the requirement into the con- text of a network system and partitioned NTCB as defined for these network criteria. NTCB protocols should be robust enough so that they permit the system to operate correctly in the case of local- ized failure. The purpose of this protection is to preserve the integrity of the NTCB itself. It is not unusual for one or more components in a network to be inoperative at any time, so it is important to minimize the effects of such failures on the rest of the network. Additional integrity and denial of service issues are addressed in Part II. IT SHOULD BE CLEAR THAT SOME INTEGRITY AND DENIAL OF SERVICE FEATURES CAN RESIDE OUTSIDE THE NTCB. OTHERWISE ALL SOFTWARE IN A NETWORK WOULD BE IN THE NTCB. EVERY PIECE OF SOFTWARE THAT HAS AN OPPORTUNITY TO WRITE TO SOME DATA OR PROTOCOL FIELD IS "TRUSTED" TO PRESERVE INTEGRITY OR NOT CAUSE DENIAL OF SERVICE TO SOME EXTENT. FOR EXAMPLE, IT IS NECESSARY TO "TRUST" TELNET TO CORRECTLY TRANSLATE USER DATA, AND TO EVENTUALLY TRANSMIT PACKETS. FTP ALSO HAS TO BE "TRUSTED" TO NOT INAPPROPRIATELY MODIFY FILES, AND TO ATTEMPT TO COMPLETE THE FILE TRANSFER. THESE PROTOCOLS CAN BE DESIGNED, HOWEVER TO EXIST OUTSIDE THE NTCB (FROM A PRO- TECTION PERSPECTIVE). IT IS BENEFICIAL TO DO THIS TYPE OF SECURITY ENGINEERING SO THAT THE AMOUNT OF CODE THAT MUST BE TRUSTED TO NOT DISCLOSE DATA IS MINIMIZED. PUTTING EVERY- THING INSIDE THE NTCB CONTRADICTS THE REQUIREMENT TO PERFORM "SIGNIFICANT SYSTEM ENGINEERING ... DIRECTED TOWARD ... EXCLUDING FROM THE TCB MODULES THAT ARE NOT PROTECTION CRIT- ICAL," WHICH REMOVES THE PRIMARY DIFFERENCE BETWEEN B2 AND B3. IF EVERYTHING HAS TO BE IN THE TCB TO ENSURE DATA INTEGRITY AND PROTECTION AGAINST DENIAL OF SERVICE, THERE WILL BE CONSIDERABLY LESS ASSURANCE THAT DISCLOSURE PROTEC- TION IS MAXIMIZED. 3.3.3.1.3 Covert Channel Analysis + Statement from DoD 5200.28-STD The system developer shall conduct a thorough search for COVERT CHANNELS and make a determination (either by actual measurement or by engineering estimation) of the maximum bandwidth of each identified channel. (See the Covert Chan- nels Guideline section.) + Interpretation The requirement, including the TCSEC Covert Channel Guideline, applies as written. In a network, there are additional instances of covert channels associated with com- munication between components. + Rationale The exploitation of network protocol information (e.g., headers) can result in covert storage channels. EXPLOITATION OF FREQUENCY OF TRANSMISSION CAN RESULT IN COVERT TIMING CHANNELS. The topic has been addressed in the literature.- 3.3.3.1.4 Trusted Facility Management + Statement from DoD 5200.28-STD The TCB shall support separate operator and administrator functions. THE FUNCTIONS PERFORMED IN THE ROLE OF A SECU- RITY ADMINISTRATOR SHALL BE IDENTIFIED. THE ADP SYSTEM ADMINISTRATIVE PERSONNEL SHALL ONLY BE ABLE TO PERFORM SECU- RITY ADMINISTRATOR FUNCTIONS AFTER TAKING A DISTINCT AUDIT- ABLE ACTION TO ASSUME THE SECURITY ADMINISTRATOR ROLE ON THE ADP SYSTEM. NON-SECURITY FUNCTIONS THAT CAN BE PERFORMED IN THE SECURITY ADMINISTRATION ROLE SHALL BE LIMITED STRICTLY TO THOSE ESSENTIAL TO PERFORMING THE SECURITY ROLE EFFEC- TIVELY. _________________________ - See, for example, Girling, C. G., "Covert Channels in LAN's," IEEE Transactions on Software Engineering, ____ ____________ __ ________ ___________ Vol. SE-13, No. 2, February 1987; and Padlipsky, M. A., Snow, D. P., and Karger, P. A., Limitations of End-to- ___________ __ ___ __ End Encryption in Secure Computer Networks, MITRE ___ __________ __ ______ ________ ________ Technical Report, MTR-3592, Vol. I, May 1978 (ESD TR 78-158, DTIC AD A059221). + Interpretation This requirement applies as written to both the network as a whole and to individual components which support such personnel. + Rationale It is recognized that based on the allocated policy elements some components may operate with no human inter- face. 3.3.3.1.5 Trusted Recovery + Statement from DoD 5200.28-STD PROCEDURES AND/OR MECHANISMS SHALL BE PROVIDED TO ASSURE THAT, AFTER AN ADP SYSTEM FAILURE OR OTHER DISCONTINUITY, RECOVERY WITHOUT A PROTECTION COMPROMISE IS OBTAINED. + Interpretation THE RECOVERY PROCESS MUST BE ACCOMPLISHED WITHOUT A PROTECTION COMPROMISE AFTER THE FAILURE OR OTHER DISCON- TINUITY OF ANY NTCB PARTITION. IT MUST ALSO BE ACCOMPLISHED AFTER A FAILURE OF THE ENTIRE NTCB. + Rationale THIS IS A STRAIGHT-FORWARD EXTENSION OF THE REQUIREMENT INTO THE NETWORK CONTEXT, AND TAKES INTO ACCOUNT THAT IT IS POSSIBLE FOR PARTS OF THE SYSTEM TO FAIL WHILE OTHER PARTS CONTINUE TO OPERATE NORMALLY. THIS MAY BE A SECURITY- RELEVANT EVENT; IF SO IT MUST BE AUDITED. 3.3.3.2 Life-Cycle Assurance 3.3.3.2.1 Security Testing + Statement from DoD 5200.28-STD The security mechanisms of the ADP system shall be tested and found to work as claimed in the system documentation. A team of individuals who thoroughly understand the specific implementation of the TCB shall subject its design documen- tation, source code, and object code to through analysis and testing. Their objectives shall be: to uncover all design and implementation flaws that would permit a subject exter- nal to the TCB to read, change, or delete data normally denied under the mandatory or discretionary security policy enforced by the TCB; as well as to assure that no subject (without authorization to do so) is able to cause the TCB to enter a state such that it is unable to respond to communications initiated by other users. The TCB shall be FOUND RESISTANT to penetration. All discovered flaws shall be removed or neutralized and the TCB retested to demon- strate that they have been eliminated and that new flaws have not been introduced. Testing shall demonstrate that the TCB implementation is consistent with the descriptive top- level specification. NO DESIGN FLAWS AND NO MORE THAN A FEW CORRECTABLE IMPLEMENTATION FLAWS MAY BE FOUND DURING TESTING AND THERE SHALL BE REASONABLE CONFIDENCE THAT FEW REMAIN. (See the Security Testing Guidelines.) + Interpretation Testing of a component will require a testbed that exercises the interfaces and protocols of the component including tests under exceptional conditions. The testing of a security mechanism of the network system for meeting this criterion shall be an integrated testing procedure involving all components containing an NTCB partition that implement the given mechanism. This integrated testing is additional to any individual component tests involved in the evaluation of the network system. The sponsor should iden- tify the allowable set of configurations including the sizes of the networks. Analysis or testing procedures and tools shall be available to test the limits of these configura- tions. A change in configuration within the allowable set of configurations does not require retesting. The testing of each component will include the intro- duction of subjects external to the NTCB partition for the component that will attempt to read, change, or delete data normally denied. If the normal interface to the component does not provide a means to create the subjects needed to conduct such a test, then this portion of the testing shall use a special version of the untrusted software for the com- ponent that results in subjects that make such attempts. The results shall be saved for test analysis. Such special versions shall have an NTCB partition that is identical to that for the normal configuration of the component under evaluation. The testing of the mandatory controls shall include tests to demonstrate that the labels for information imported and/or exported to/from the component accurately represent the labels maintained by the NTCB partition for the component for use as the basis for its mandatory access control decisions. The tests shall include each type of device, whether single-level or multilevel, supported by the component. The NTCB must be FOUND RESISTANT to penetration. This applies to the NTCB as a whole, and to each NTCB partition in a component of this class. + Rationale The phrase "no subject (without authorization to do so) is able to cause the TCB to enter a state such that it is unable to respond to communications initiated by other users" relates to the security services (Part II of this TNI) for the Denial of Service problem, and to correctness of the protocol implementations. Testing is an important method available in this evaluation division to gain any assurance that the security mechanisms perform their intended function. A major purpose of testing is to demonstrate the system's response to inputs to the NTCB partition from untrusted (and possibly mali- cious) subjects. In contrast to general purpose systems that allow for the dynamic creation of new programs and the introductions of new processes (and hence new subjects) with user speci- fied security properities, many network components have no method for introducing new programs and/or processes during their normal operation. Therefore, the programs necessary for the testing must be introduced as special versions of the software rather than as the result of normal inputs by the test team. However, it must be insured that the NTCB partition used for such tests is identical to the one under evaluation. Sensitivity labels serve a critical role in maintaining the security of the mandatory access controls in the net- work. Especially important to network security is the role of the labels for information communicated between com- ponents - explicit labels for multilevel devices and impli- cit labels for single-level devices. Therefore the testing for correct labels is highlighted. The requirement for testing to demonstrate consistency between the NTCB implementation and the DTLS is a straight- forward extension of the TCSEC requirement into the context of a network system. 3.3.3.2.2 Design Specification and Verification + Statement from DoD 5200.28-STD A formal model of the security policy supported by the TCB shall be maintained over the life cycle of the ADP system that is proven and demonstrated to be consistent with its axioms. A descriptive top-level specification (DTLS) of the TCB shall be maintained that completely and accurately describes the TCB in terms of exceptions, error messages, and effects. It shall be shown to be an accurate descrip- tion of the TCB interface. A CONVINCING ARGUMENT SHALL BE GIVEN THAT THE DTLS IS CONSISTENT WITH THE MODEL. + Interpretation The overall network security policy expressed in this model will provide the basis for the mandatory access con- trol policy exercised by the NTCB over subjects and storage objects in the entire network. The policy will also be the basis for the discretionary access control policy exercised by the NTCB to control access of named users to named objects. Data integrity requirements addressing the effects of unauthorized MSM need not be included in this model. The overall network policy must be decomposed into policy ele- ments that are allocated to appropriate components and used as the basis for the security policy model for those com- ponents. The level of abstraction of the model, and the set of subjects and objects that are explicitly represented in the model, will be affected by the NTCB partitioning. Subjects and objects must be represented explicitly in the model for the partition if there is some network component whose NTCB partition exercises access control over them. The model shall be structured so that the axioms and entities applica- ble to individual network components are manifest. Global network policy elements that are allocated to components shall be represented by the model for that component. The requirements for a network DTLS are given in the Design Documentation section. + Rationale The treatment of the model depends to a great extent on the degree of integration of the communications service into a distributed system. In a closely coupled distributed sys- tem, one might use a model that closely resembles one appropriate for a stand-alone computer system. In ALL cases, the model of each partition will be expected to show the role of the NTCB partition in each kind of component. It will most likely clarify the model, although not part of the model, to show access restrictions implied by the system design; for example, subjects representing protocol entities might have access only to objects containing data units at the same layer of protocol. The allocation of subjects and objects to different proto- col layers is a protocol design choice which need not be reflected in the security policy model. 3.3.3.2.3 Configuration Management + Statement from DoD 5200.28-STD During development and maintenance of the TCB, a configura- tion management system shall be in place that maintains con- trol of changes to the descriptive top-level specification, other design data, implementation documentation, source code, the running version of the object code, and test fix- tures and documentation. The configuration management sys- tem shall assure a consistent mapping among all documenta- tion and code associated with the current version of the TCB. Tools shall be provided for generation of a new ver- sion of the TCB from source code. Also available shall be tools for comparing a newly generated version with the pre- vious TCB version in order to ascertain that only the intended changes have been made in the code that will actu- ally be used as the new version of the TCB. + Interpretation The requirement applies as written, with the following extensions: 1. A configuration management system must be in place for each NTCB partition. 2. A configuration management plan must exist for the entire system. If the configuration management sys- tem is made up of the conglomeration of the confi- guration management systems of the various NTCB par- titions, then the configuration management plan must address the issue of how configuration control is applied to the system as a whole. + Rationale Each NTCB partition must have a configuration manage- ment system in place, or else there will be no way for the NTCB as a whole to have an effective configuration manage- ment system. The other extensions are merely reflections of the way that networks operate in practice. 3.3.4 Documentation. _ _ _ _____________ 3.3.4.1 Security Features User's Guide + Statement from DoD 5200.28-STD A single summary, chapter, or manual in user documentation shall describe the protection mechanisms provided by the TCB, interpretations on their use, and how they interact with one another. + Interpretation This user documentation describes user visible protec- tion mechanisms at the global (network system) level and at the user interface of each component, and the interaction among these. + Rationale The interpretation is an extension of the requirement into the context of a network system as defined for these network criteria. Documentation of protection mechanisms provided by individual components is required by the cri- teria for trusted computer systems that are applied as appropriate for the individual components. 3.3.4.2 Trusted Facility Manual + Statement from DoD 5200.28-STD A manual addressed to the ADP system administrator shall present cautions about functions and privileges that should be controlled when running a secure facility. The procedures for examining and maintaining the audit files as well as the detailed audit record structure for each type of audit event shall be given. The manual shall describe the operator and administrator functions related to security, to include changing the security characteristics of a user. It shall provide interpretations on the consistent and effective use of the protection features of the system, how they interact, how to securely generate a new TCB, and facility procedures, warnings, and privileges that need to be controlled in order to operate the facility in a secure manner. The TCB modules that contain the reference validation mechanism shall be identified. The procedures for secure generation of a new TCB from source after modification of any modules in the TCB shall be described. IT SHALL INCLUDE THE PROCEDURES TO ENSURE THAT THE SYSTEM IS INITIALLY STARTED IN A SECURE MANNER. PROCEDURES SHALL ALSO BE INCLUDED TO RESUME SECURE SYSTEM OPERATION AFTER ANY LAPSE IN SYSTEM OPERATION. + Interpretation This manual shall contain specifications and procedures to assist the system administrator(s) maintain cognizance of the network configuration. These specifications and pro- cedures shall address the following: 1. The hardware configuration of the network itself; 2. The implications of attaching new components to the network; 3. The case where certain components may periodically leave the network (e.g., by crashing, or by being disconnected) and then rejoin; 4. Network configuration aspects that can impact the security of the network system; (For example, the manual should describe for the network system administrator the interconnections among components that are consistent with the overall network system architecture.) 5. Loading or modifying NTCB software or firmware (e.g., down-line loading). 6. Incremental updates; that is, it must explicitly indicate which components of the network may change without others also changing. The physical and administrative environmental controls shall be specified. Any assumptions about security of a given network should be clearly stated (e.g., the fact that all communications links must be physically protected to a certain level). The components of the network that form the NTCB must be identified. Furthermore, the modules within an NTCB par- tition that contain the reference validation mechanism (if any) within that partition must be identified. The procedures for the secure generation of a new ver- sion (or copy) of each NTCB partition from source must be described. The procedures and requirements for the secure generation of the NTCB necessitated by changes in the net- work configuration shall be described. PROCEDURES FOR STARTING EACH NTCB PARTITION IN A SECURE STATE SHALL BE SPECIFIED. PROCEDURES MUST ALSO BE INCLUDED TO RESUME SECURE OPERATION OF EACH NTCB PARTITION AND/OR THE NTCB AFTER ANY LAPSE IN SYSTEM OR SUBSYSTEM OPERATION. + Rationale There may be multiple system administrators with diverse responsibilities. The technical security measures described by these criteria must be used in conjunction with other forms of security in order to achieve security of the network. Additional forms include administrative security, physical security, emanations security, etc. Extension of this criterion to cover configuration aspects of the network is needed because, for example, proper interconnection of components is typically essential to achieve a correct realization of the network architec- ture. As mentioned in the section on Label Integrity, cryp- tography is one common mechanism employed to protect commun- ication circuits. Encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity of the ciphertext is generally lower than the cleartext. If encryption methodologies are employed, they shall be approved by the National Security Agency (NSA). The encryption algorithm and its implementation are outside the scope of these interpretations. This algorithm and implementation may be implemented in a separate device or may be a function of a subject in a component not dedi- cated to encryption. Without prejudice, either implementa- tion packaging is referred to as an encryption mechanism herein. The requirements for descriptions of NTCB generation and identification of modules and components that form the NTCB are straightforward extensions of the TCSEC require- ments into the network context. In those cases where the vendor does not provide source code, an acceptable procedure shall be to request the vendor to perform the secure genera- tion. GIVEN THE NATURE OF NETWORK SYSTEMS (E.G., VARIOUS COM- PONENTS TEND TO BE DOWN AT DIFFERENT TIMES, AND THE NETWORK SYSTEM MUST CONTINUE OPERATION WITHOUT THAT COMPONENT), IT IS IMPERATIVE TO KNOW BOTH HOW TO SECURELY START UP AN NTCB PARTITION, AND HOW TO RESUME OPERATION SECURELY. IT IS ALSO NECESSARY TO KNOW HOW TO RESUME SECURE OPERATION OF THE NTCB AFTER ANY PARTITION HAS BEEN DOWN. 3.3.4.3 Test Documentation + Statement from DoD 5200.28-STD The system developer shall provide to the evaluators a docu- ment that describes the test plan, test procedures that show how the security mechanisms were tested, and results of the security mechanisms' functional testing. It shall include results of testing the effectiveness of the methods used to reduce covert channel bandwidths. + Interpretation The "system developer" is interpreted as "the network system sponsor". The description of the test plan should establish the context in which the testing was or should be conducted. The description should identify any additional test components that are not part of the system being evaluated. This includes a description of the test-relevant functions of such test components and a description of the interfacing of those test components to the system being evaluated. The description of the test plan should also demonstrate that the tests adequately cover the network security policy. The tests should include the features described in the System Architecture and the System Integrity sections. The tests should also include network configuration and sizing. + Rationale The entity being evaluated may be a networking subsys- tem (see Appendix A) to which other components must be added to make a complete network system. In that case, this interpretation is extended to include contextual definition because, at evaluation time, it is not possible to validate the test plans without the description of the context for testing the networking subsystem. The bandwidths of covert channels are used to determine the suitability of a network system for a given environment. The effectiveness of the methods used to reduce these bandwidths must therefore be accurately determined. 3.3.4.4 Design Documentation + Statement from DoD 5200.28-STD Documentation shall be available that provides a description of the manufacturer's philosophy of protection and an expla- nation of how this philosophy is translated into the TCB. The interfaces between the TCB modules shall be described. A formal description of the security policy model enforced by the TCB shall be available and an explanation provided to show that it is sufficient to enforce the security policy. The specific TCB protection mechanisms shall be identified and an explanation given to show that they satisfy the model. The descriptive top-level specification (DTLS) shall be shown to be an accurate description of the TCB interface. Documentation shall describe how the TCB implements the reference monitor concept and give an explanation why it is tamper resistant, cannot be bypassed, and is correctly implemented. THE TCB IMPLEMENTATION (I.E., IN HARDWARE, FIRMWARE, AND SOFTWARE) SHALL BE INFORMALLY SHOWN TO BE CON- SISTENT WITH THE DTLS. THE ELEMENTS OF THE DTLS SHALL BE SHOWN, USING INFORMAL TECHNIQUES, TO CORRESPOND TO THE ELE- MENTS OF THE TCB. Documentation shall describe how the TCB is structured to facilitate testing and to enforce least privilege. This documentation shall also present the results of the covert channel analysis and the tradeoffs involved in restricting the channels. All auditable events that may be used in the exploitation of known covert storage channels shall be identified. The bandwidths of known covert storage channels, the use of which is not detectable by the auditing mechanisms, shall be provided. (See the Covert Channel Guideline section.) + Interpretation Explanation of how the sponsor's philosophy of protec- tion is translated into the NTCB shall include a description of how the NTCB is partitioned. The security policy also shall be stated. The description of the interfaces between the NTCB modules shall include the interface(s) between NTCB partitions and modules within the partitions if the modules exist. The sponsor shall describe the security architecture and design, including the allocation of security require- ments among components. The documentation includes both a system description and a set of component DTLS's. The system description addresses the network security architecture and design by specifying the types of components in the network, which ones are trusted, and in what way they must cooperate to support network security objectives. A component DTLS shall be provided for each trusted network component, i.e., each component containing an NTCB partition. Each component DTLS shall describe the interface to the NTCB partition of its component. BOTH THE SYSTEM DESCRIPTION AND EACH COMPONENT DTLS SHALL BE SHOWN CONSISTENT WITH THOSE ASSERTIONS IN THE MODEL THAT APPLY TO IT. Appendix A addresses component evaluation issues. TO SHOW THE CORRESPONDENCE BETWEEN THE DTLS AND THE NTCB IMPLEMENTATION, IT SUFFICES TO SHOW CORRESPONDENCE BETWEEN EACH COMPONENT DTLS AND THE NTCB PARTITION IN THAT COMPONENT. As stated in the introduction to Division B, the spon- sor must demonstrate that the NTCB employs the reference monitor concept. The security policy model must be a model for a reference monitor. The security policy model for each partition implement- ing a reference monitor shall fully represent the access control policy supported by the partition, including the discretionary and mandatory security policy for secrecy and/or integrity. For the mandatory policy the single domi- nance relation for sensitivity labels, including secrecy and/or integrity components, shall be precisely defined. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system as defined for this network interpretation. Other documenta- tion, such as description of components and description of operating environment(s) in which the networking subsystem or network system is designed to function, is required else- where, e.g., in the Trusted Facility Manual. In order to be evaluated, a network must possess a coherent Network Security Architecture and Design. (Inter- connection of components that do not adhere to such a single coherent Network Security Architecture is addressed in the Interconnection of Accredited AIS, Appendix C.) The Network Security Architecture must address the security-relevant policies, objectives, and protocols. The Network Security Design specifies the interfaces and services that must be incorporated into the network so that it can be evaluated as a trusted entity. There may be multiple designs that con- form to the same architecture but are more or less incompa- tible and non-interoperable (except through the Interconnec- tion Rules). Security related mechanisms requiring coopera- tion among components are specified in the design in terms of their visible interfaces; mechanisms having no visible interfaces are not specified in this document but are left as implementation decisions. The Network Security Architecture and Design must be available from the network sponsor before evaluation of the network, or any component, can be undertaken. The Network Security Architecture and Design must be sufficiently com- plete, unambiguous, and free from obvious flaws to permit the construction or assembly of a trusted network based on the structure it specifies. When a component is being designed or presented for evaluation, or when a network assembled from components is assembled or presented for evaluation, there must be a priori evidence that the Network security Architecture and Design are satisfied. That is, the components can be assem- bled into a network that conforms in every way with the Net- work Security Architecture and Design to produce a physical realization that is trusted to the extent that its evalua- tion indicates. In order for a trusted network to be constructed from components that can be built independently, the Network Security Architecture and Design must completely and unambi- guously define the security functionality of components as well as the interfaces between or among components. The Network Security Architecture and Design must be evaluated to determine that a network constructed to its specifica- tions will in fact be trusted, that is, it will be evaluat- able under these interpretations. The term "model" is used in several different ways in a network context, e.g., a "protocol reference model," a "for- mal network model," etc. Only the "security policy model" is addressed by this requirement and is specifically intended to model the interface, viz., "security perimeter," of the reference monitor and must meet all the requirements defined in the TCSEC. It must be shown that all parts of the TCB are a valid interpretation of the security policy model, i.e., that there is no change to the secure state except as represented by the model. 4.0 DIVISION A: VERIFIED PROTECTION _ _ ________ _ ________ __________ This division is characterized by the use of formal security methods to assure that the mandatory and discretionary secu- rity controls employed in the network system can effectively protect classified or other sensitive information stored or processed by the system. Extensive documentation is re- quired to demonstrate that the NTCB meets the security re- quirements in all aspects of design, development and imple- mentation. 4.1 CLASS (A1): VERIFIED DESIGN _ _ _____ __ ________ ______ SYSTEMS IN CLASS (A1) ARE FUNCTIONALLY EQUIVALENT TO THOSE IN CLASS (B3) IN THAT NO ADDITIONAL ARCHITECTURAL FEATURES OR POLICY REQUIREMENTS ARE ADDED. THE DISTINGUISHING FEATURE OF SYSTEMS IN THIS CLASS IS THE ANALYSIS DERIVED FROM FORMAL DESIGN SPECIFICATION AND VERIFICATION TECHNIQUES AND THE RESULTING HIGH DEGREE OF ASSURANCE THAT THE NTCB IS CORRECTLY IMPLEMENTED. THIS ASSURANCE IS DEVELOPMENTAL IN NATURE, STARTING WITH A FORMAL MODEL OF THE SECURITY POLICY AND A FORMAL TOP- LEVEL SPECIFICATION (FTLS) OF THE DESIGN. INDEPENDENT OF THE PARTICULAR SPECIFICATION LANGUAGE OR VERIFICATION SYSTEM USED, THERE ARE FIVE IMPORTANT CRITERIA FOR CLASS (A1) DESIGN VERIFICATION: + A FORMAL MODEL OF THE SECURITY POLICY MUST BE CLEARLY IDENTIFIED AND DOCUMENTED, INCLUDING A MATHEMATICAL PROOF THAT THE MODEL IS CONSISTENT WITH ITS AXIOMS AND IS SUFFICIENT TO SUPPORT THE SECURITY POLICY. + AN FTLS MUST BE PRODUCED THAT INCLUDES ABSTRACT DEFINITIONS OF THE FUNCTIONS THE NTCB PERFORMS AND OF THE HARDWARE AND/OR FIRMWARE MECHANISMS THAT ARE USED TO SUPPORT SEPARATE EXECUTION DOMAINS. + THE FTLS OF THE NTCB MUST BE SHOWN TO BE CON- SISTENT WITH THE MODEL BY FORMAL TECHNIQUES WHERE POSSIBLE (I.E., WHERE VERIFICATION TOOLS EXIST) AND INFORMAL ONES OTHERWISE. + THE NTCB IMPLEMENTATION (I.E., IN HARDWARE, FIRMWARE, AND SOFTWARE) MUST BE INFORMALLY SHOWN TO BE CONSISTENT WITH THE FTLS. THE ELEMENTS OF THE FTLS MUST BE SHOWN, USING INFORMAL TECH- NIQUES, TO CORRESPOND TO THE ELEMENTS OF THE NTCB. THE FTLS MUST EXPRESS THE UNIFIED PROTEC- TION MECHANISM REQUIRED TO SATISFY THE SECURITY POLICY, AND IT IS THE ELEMENTS OF THIS PROTECTION MECHANISM THAT ARE MAPPED TO THE ELEMENTS OF THE NTCB. + FORMAL ANALYSIS TECHNIQUES MUST BE USED TO IDEN- TIFY AND ANALYZE COVERT CHANNELS. INFORMAL TECH- NIQUES MAY BE USED TO IDENTIFY COVERT TIMING CHANNELS. THE CONTINUED EXISTENCE OF IDENTIFIED COVERT CHANNELS IN THE SYSTEM MUST BE JUSTIFIED. IN KEEPING WITH THE EXTENSIVE DESIGN AND DEVELOP- MENT ANALYSIS OF THE NTCB REQUIRED OF SYSTEMS IN CLASS (A1), MORE STRINGENT CONFIGURATION MANAGE- MENT IS REQUIRED AND PROCEDURES ARE ESTABLISHED FOR SECURELY DISTRIBUTING THE SYSTEM TO SITES. A SYSTEM SECURITY ADMINISTRATOR IS SUPPORTED. THE FOLLOWING ARE MINIMAL REQUIREMENTS FOR SYSTEM ASSIGNED A CLASS (A1) RATING: 4.1.1 Security Policy _ _ _ ________ ______ + Statement from DoD 5200.28-STD Implied from the Introduction to the TCSEC. + Interpretation The network sponsor shall describe the overall network security policy enforced by the NTCB. At a minimum, this policy shall include the discretionary and mandatory requirements applicable to this class. The policy may require data secrecy, or data integrity, or both. The pol- icy is an access control policy having two primary com- ponents: mandatory and discretionary. The policy shall include a discretionary policy for protecting the informa- tion being processed based on the authorizations of indivi- duals, users, or groups of users. This access control pol- icy statement shall describe the requirements on the network to prevent or detect "reading or destroying" sensitive information by unauthorized users or errors. The mandatory policy must define the set of distinct sensitivity levels that it supports. For the Class B1 or above the mandatory policy shall be based on the labels associated with the information that reflects its sensitivity with respect to secrecy and/or integrity, where applicable, and labels asso- ciated with users to reflect their authorization to access such information. Unauthorized users include both those that are not authorized to use the network at all (e.g., a user attempting to use a passive or active wire tap) or a legitimate user of the network who is not authorized to access a specific piece of information being protected. Note that "users" does not include "operators," "system programmers," "technical control officers," "system security officers," and other system support personnel. They are distinct from users and are subject to the Trusted Facility Manual and the System Architecture requirements. Such indi- viduals may change the system parameters of the network sys- tem, for example, by defining membership of a group. These individuals may also have the separate role of users. SECRECY POLICY: The network sponsor shall define the form of the discretionary and mandatory secrecy policy that is enforced in the network to prevent unauthorized users from reading the sensitive infor- mation entrusted to the network. DATA INTEGRITY POLICY: The network sponsor shall define the discretionary and mandatory integrity policy to prevent unauthorized users from modifying, viz., writing, sensitive information. The defini- tion of data integrity presented by the network sponsor refers to the requirement that the informa- tion has not been subjected to unauthorized modifi- cation in the network. The mandatory integrity pol- icy enforced by the NTCB cannot, in general, prevent modification while information is being transmitted between components. However, an integrity sensi- tivity label may reflect the confidence that the information has not been subjected to transmission errors because of the protection afforded during transmission. This requirement is distinct from the requirement for label integrity. + Rationale The word "sponsor" is used in place of alternatives (such as "vendor," "architect," "manufacturer," and "developer") because the alternatives indicate people who may not be available, involved, or relevant at the time that a network system is proposed for evaluation. A trusted network is able to control both the reading and writing of shared sensitive information. Control of writing is used to protect against destruction of informa- tion. A network normally is expected to have policy require- ments to protect both the secrecy and integrity of the information entrusted to it. In a network the integrity is frequently as important or more important than the secrecy requirements. Therefore the secrecy and/or integrity policy to be enforced by the network must be stated for each net- work regardless of its evaluation class. The assurance that the policy is faithfully enforced is reflected in the evaluation class of the network. This control over modification is typically used to protect information so that it may be relied upon and to control the potential harm that would result if the informa- tion were corrupted. The overall network policy require- ments for integrity includes the protection for data both while being processed in a component and while being transmitted in the network. The access control policy enforced by the NTCB relates to the access of subjects to objects within each component. Communications integrity addressed within Part II relates to information while being transmitted. The mandatory integrity policy (at class B1 and above) in some architectures may be useful in supporting the link- age between the connection oriented abstraction introduced in the Introduction and the individual components of the network. For example, in a key distribution center for end-to-end encryption, a distinct integrity category may be assigned to isolate the key generation code and data from possible modification by other supporting processes in the same component, such as operator interfaces and audit. The mandatory integrity policy for some architecture may define an integrity sensitivity label that reflects the specific requirements for ensuring that information has not been subject to random errors in excess of a stated limit nor to unauthorized message stream modification (MSM) -. The specific metric associated with an integrity sensitivity label will generally reflect the intended applications of the network. 4.1.1.1 Discretionary Access Control + Statement from DoD 5200.28-STD The TCB shall define and control access between named users and named objects (e.g., files and programs) in the ADP sys- tem. The enforcement mechanism (e.g., access control lists) shall allow users to specify and control sharing of those objects and shall provide controls to limit propagation of access rights. The discretionary access control mechanism shall, either by explicit user action or by default, provide that objects are protected from unauthorized access. These access controls shall be capable of specifying, for each named object, a list of named individuals and a list of groups of named individuals with their respective modes of access to that object. Furthermore, for each such named object, it shall be possible to specify a list of named individuals and a list of groups of named individuals for which no access to the object is given. Access permission to an object by users not already possessing access _________________________ - See Voydock, Victor L. and Stephen T. Kent, "Secu- rity Mechanisms in High-Level Network Protocols," Com- ___ puting Surveys, Vol. 15, No. 2, June 1983, pp 135-171. ______ _______ permission shall only be assigned by authorized users. + Interpretation The discretionary access control (DAC) mechanism(s) may be distributed over the partitioned NTCB in various ways. Some part, all, or none of the DAC may be implemented in a given component of the network system. In particular, com- ponents that support only internal subjects (i.e., that have no subjects acting as direct surrogates for users), such as a public network packet switch, might not implement the DAC mechanism(s) directly (e.g., they are unlikely to contain access control lists). Identification of users by groups may be achieved in various ways in the networking environment. For example, the network identifiers (e.g., internet addresses) for vari- ous components (e.g., hosts, gateways) can be used as iden- tifiers of groups of individual users (e.g., "all users at Host A," "all users of network Q") so long as the individu- als involved in the group are implied by the group identif- ier. For example, Host A might employ a particular group-id, for which it maintains a list of explicit users in that group, in its network exchange with Host B, which accepts the group-id under the conditions of this interpretation. For networks, individual hosts will impose need-to-know controls over their users on the basis of named individuals - much like (in fact, probably the same) controls used when there is no network connection. When group identifiers are acceptable for access con- trol, the identifier of some other host may be employed, to eliminate the maintenance that would be required if indivi- dual identification of remote users was employed. In class C2 and higher, however, it must be possible from that audit record to identify (immediately or at some later time) exactly the individuals represented by a group identifier at the time of the use of that identifier. There is allowed to be an uncertainty because of elapsed time between changes in the group membership and the enforcement in the access con- trol mechanisms. The DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. The reference monitor manages all the physical resources of the system and from them creates the abstraction of subjects and objects that it con- trols. Some of these subjects and objects may be used to implement a part of the NTCB. When the DAC mechanism is distributed in such NTCB subjects (i.e., when outside the reference monitor), the assurance requirements (see the Assurance section) for the design and implementation of the DAC shall be those of class C2 for all networks of class C2 or above. When integrity is included as part of the network dis- cretionary security policy, the above interpretations shall be specifically applied to the controls over modification, viz, the write mode of access, within each component based on identified users or groups of users. + Rationale In this class, the supporting elements of the overall DAC mechanism are required to isolate information (objects) that supports DAC so that it is subject to auditing require- ments (see the System Architecture section). The use of network identifiers to identify groups of individual users could be implemented, for example, as an X.25 community of interest in the network protocol layer (layer 3). In all other respects, the supporting elements of the overall DAC mechanism are treated exactly as untrusted subjects are treated with respect to DAC in an ADP system, with the same result as noted in the interpretation. A typical situation for DAC is that a surrogate process for a remote user will be created in some host for access to objects under the control of the NTCB partition within that host. The interpretation requires that a user identifier be assigned and maintained for each such process by the NTCB, so that access by a surrogate process is subject to essen- tially the same discretionary controls as access by a pro- cess acting on behalf of a local user would be. However, within this interpretation a range of possible interpreta- tions of the assigned user identification is permitted. The most obvious situation would exist if a global database of network users were to be made permanently avail- able on demand to every host, (i.e., a name server existed) so that all user identifications were globally meaningful. It is also acceptable, however, for some NTCB parti- tions to maintain a database of locally-registered users for its own use. In such a case, one could choose to inhibit the creation of surrogate processes for locally unregistered users, or (if permitted by the local policy) alternatively, to permit the creation of surrogate processes with preselected user and group identifiers which, in effect, identify the process as executing on behalf of a member of a group of users on a particular remote host. The intent of the words concerning audit in the interpretation is to pro- vide a minimally acceptable degree of auditability for cases such as the last described. What is required is that there be a capability, using the audit facilities provided by the network NTCB partitions involved, to determine who was logged in at the actual host of the group of remote users at the time the surrogate processing occured. Associating the proper user id with a surrogate process is the job of identification and authentication. This means that DAC is applied locally, with respect to the user id of the surrogate process. The transmission of the data back across the network to the user's host, and the creation of a copy of the data there, is not the business of DAC. Components that support only internal subjects impact the implementation of the DAC by providing services by which information (e.g., a user-id) is made available to a com- ponent that makes a DAC decision. An example of the latter would be the case that a user at Host A attempts to access a file at Host B. The DAC decision might be (and usually would be) made at Host B on the basis of a user-id transmit- ted from Host A to Host B. Unique user identification may be achieved by a variety of mechanisms, including (a) a requirement for unique iden- tification and authentication on the host where access takes place; (b) recognition of fully qualified network addresses authenticated by another host and forwarded to the host where access takes place; or (c) administrative support of a network-wide unique personnel identifier that could be authenticated and forwarded by another host as in (b) above, or could be authenticated and forwarded by a dedicated net- work identification and authentication server. The proto- cols which implement (b) or (c) are subject to the System Architecture requirements. Network support for DAC might be handled in other ways than that described as "typical" above. In particular, some form of centralized access control is often proposed. An access control center may make all decisions for DAC, or it may share the burden with the hosts by controlling host-to- host connections, and leaving the hosts to decide on access to their objects by users at a limited set of remote hosts. In this case the access control center provides the linkage between the connection oriented abstraction (as discussed in the Introduction) and the overall network security policy for DAC. In all cases the enforcement of the decision must be provided by the host where the object resides. There are two forms of distribution for the DAC mechan- ism: implementing portions of the DAC in separate com- ponents, and supporting the DAC in subjects contained within the NTCB partition in a component. Since "the ADP system" is understood to be "the computer network" as a whole, each network component is responsible for enforcing security in the mechanisms allocated to it to ensure secure implementa- tion of the network security policy. For traditional host systems it is frequently easy to also enforce the DAC along with the MAC within the reference monitor, per se, although a few approaches, such as virtual machine monitors, support DAC outside this interface. In contrast to the universally rigid structure of man- datory policies (see the Mandatory Access Control section), DAC policies tend to be very network and system specific, with features that reflect the natural use of the system. For networks it is common that individual hosts will impose controls over their local users on the basis of named individuals-much like the controls used when there is no network connection. However, it is difficult to manage in a centralized manner all the individuals using a large net- work. Therefore, users on other hosts are commonly grouped together so that the controls required by the network DAC policy are actually based on the identity of the hosts or other components. A gateway is an example of such a com- ponent. The assurance requirements are at the very heart of the concept of a trusted system. It is the assurance that determines if a system or network is appropriate for a given environment, as reflected, for example, in the Environments Guideline-. In the case of monolithic systems that have DAC integral to the reference monitor, the assurance require- ments for DAC are inseparable from those of the rest of the reference monitor. For networks there is typically a much clearer distinction due to distributed DAC. The rationale for making the distinction in this network interpretation is that if major trusted network components can be made signi- ficantly easier to design and implement without reducing the ability to meet security policy, then trusted networks will be more easily available. 4.1.1.2 Object Reuse + Statement from DoD 5200.28-STD All authorizations to the information contained within a storage object shall be revoked prior to initial assignment, allocation or reallocation to a subject from the TCB's pool of unused storage objects. No information, including encrypted representations of information, produced by a prior subject's actions is to be available to any subject that obtains access to an object that has been released back to the system. + Interpretation The NTCB shall ensure that any storage objects that it controls (e.g., message buffers under the control of a NTCB partition in a component) contain no information for which a subject in that component is not authorized before granting access. This requirement must be enforced by each of the NTCB partitions. _________________________ - Guidance for Applying the Department of Defense ________ ___ ________ ___ __________ __ _______ Trusted Computer System Evaluation Criteria in Specific _______ ________ ______ __________ ________ __ ________ Environments, CSC-STD-003-85. ____________ + Rationale In a network system, storage objects of interest are things that the NTCB directly controls, such as message buffers in components. Each component of the network system must enforce the object reuse requirement with respect to the storage objects of interest as determined by the network security policy. For example, the DAC requirement in this division leads to the requirement here that message buffers be under the control of the NTCB partition. A buffer assigned to an internal subject may be reused at the discre- tion of that subject which is responsible for preserving the integrity of message streams. Such controlled objects may be implemented in physical resources, such as buffers, disk sectors, tape space, and main memory, in components such as network switches. 4.1.1.3 Labels + Statement from DoD 5200.28-STD Sensitivity labels associated with each ADP system resource (e.g., subject, storage object, ROM) that is directly or indirectly accessible by subjects external to the TCB shall be maintained by the TCB. These labels shall be used as the basis for mandatory access control decisions. In order to import non-labeled data, the TCB shall request and receive from an authorized user the sensitivity level of the data, and all such actions shall be auditable by the TCB. + Interpretation Non-labeled data imported under the control of the NTCB partition will be assigned a label constrained by the device labels of the single-level device used to import it. Labels may include secrecy and integrity- components in accordance with the overall network security policy described by the network sponsor. Whenever the term "label" is used throughout this interpretation, it is understood to include both components as applicable. Similarly, the terms "single-level" and "multilevel" are understood to be based on both the secrecy and integrity components of the policy. The mandatory integrity policy will typically have require- ments, such as the probability of undetected message stream modification, that will be reflected in the label for the data so protected. For example, when data is imported its integrity label may be assigned based on mechanisms, such as cryptography, used to provide the assurance required by the policy. The NTCB shall assure that such mechanism are pro- tected from tampering and are always invoked when they are _________________________ - See, for example, Biba, K.J., "Integrity Considera- tion for Secure Computer Systems," ESD-TR-76-372, MTR- 3153, The MITRE Corporation, Bedford, MA, April 1977. the basis for a label. If the security policy includes an integrity policy, all activities that result in message-stream modification during transmission are regarded as unauthorized accesses in violation of the integrity policy. The NTCB shall have an automated capability for testing, detecting, and reporting those errors/corruptions that exceed specified network integrity policy requirements. Message-stream modification (MSM) countermeasures shall be identified. A technology of adequate strength shall be selected to resist MSM. If encryption methodologies are employed, they shall be approved by the National Security Agency. All objects must be labeled within each component of the network that is trusted to maintain separation of multi- ple levels of information. The label associated with any objects associated with single-level components will be identical to the level of that component. Objects used to store network control information, and other network struc- tures, such as routing tables, must be labeled to prevent unauthorized access and/or modification. + Rationale The interpretation is an extension of the requirement into the context of a network system and partitioned NTCB as defined for these network interpretations. A single-level device may be regarded either as a subject or an object. A multilevel device is regarded as a trusted subject in which the security range of the subject is the minimum-maximum range of the data expected to be transmitted over the dev- ice. The sensitivity labels for either secrecy or integrity or both may reflect non-hierarchical categories or hierarch- ical classification or both. For a network it is necessary that this requirement be applied to all network system resources at the (B2) level and above. The NTCB is responsible for implementing the network integrity policy, when one exists. The NTCB must enforce that policy by ensuring that information is accurately transmitted from source to destination (regardless of the number of intervening connecting points). The NTCB must be able to counter equipment failure, environmental disrup- tions, and actions by persons and processes not authorized to alter the data. Protocols that perform code or format conversion shall preserve the integrity of data and control information. The probability of an undetected transmission error may be specified as part of the network security policy so that the acceptability of the network for its intended application may be determined. The specific metrics (e.g., probability of undetected modification) satisfied by the data can be reflected in the integrity sensitivity label associated with the data while it is processed within a com- ponent. It is recognized that different applications and operational environments (e.g., crisis as compared to logis- tic) will have different integrity requirements. The network shall also have an automated capability of testing for, detecting, and reporting errors that exceed a threshold consistent with the operational mode requirements. The effectiveness of integrity countermeasures must be esta- blished with the same rigor as the other security-relevant properties such as secrecy. Cryptography is often utilized as a basis to provide data integrity assurance. Mechanisms, such as Manipulation Detection Codes (MDC)-, may be used. The adequacy of the encryption or MDC algorithm, the correctness of the protocol logic, and the adequacy of implementation must be esta- blished in MSM countermeasures design. 4.1.1.3.1 Label Integrity + Statement from DoD 5200.28-STD Sensitivity labels shall accurately represent sensitivity levels of the specific subjects or objects with which they are associated. When exported by the TCB, sensitivity labels shall accurately and unambiguously represent the internal labels and shall be associated with the information being exported. + Interpretation The phrase "exported by the TCB" is understood to include transmission of information from an object in one component to an object in another component. Information transferred between NTCB partitions is addressed in the Sys- tem Integrity Section. The form of internal and external (exported) sensitivity labels may differ, but the meaning shall be the same. The NTCB shall, in addition, ensure that correct association of sensitivity labels with the informa- tion being transported across the network is preserved. As mentioned in the Trusted Facility Manual Section, encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity level of the ciphertext is generally lower than the cleartext. It fol- lows that cleartext and ciphertext are contained in _________________________ - See Jueneman, R. R., "Electronic Document Authenti- cation," IEEE Network Magazine, April 1987, pp 17-23. ____ _______ ________ different objects, each possessing its own label. The label of the cleartext must be preserved and associated with the ciphertext so that it can be restored when the cleartext is subsequently obtained by decrypting the ciphertext. If the cleartext is associated with a single-level device, the label of that cleartext may be implicit. The label may also be implicit in the key. When information is exported to an environment where it is subject to deliberate or accidental modification, the TCB shall support the means, such as cryptographic checksums, to assure the accuracy of the labels. When there is a manda- tory integrity policy, the policy will define the meaning of integrity labels. + Rationale Encryption algorithms and their implementation are out- side the scope of these interpretations. Such algorithms may be implemented in a separate device or may be incor- porated in a subject of a larger component. Without preju- dice, either implementation packaging is referred to as an encryption mechanism herein. If encryption methodologies are employed in this regard, they shall be approved by the National Security Agency (NSA). The encryption process is part of the Network Trusted Computer Base partition in the components in which it is implemented. The encryption mechanism is not necessarily a mul- tilevel device or multilevel subject, as these terms are used in these criteria. The process of encryption is mul- tilevel by definition. The cleartext and ciphertext inter- faces carry information of different sensitivity. An encryption mechanism does not process data in the sense of performing logical or arithmetic operations on that data with the intent of producing new data. The cleartext and ciphertext interfaces on the encryption mechanism must be separately identified as being single-level or multilevel. If the interface is single-level, then the sensitivity of the data is established by a trusted individual and impli- citly associated with the interface; the Exportation to Single-Level Devices criterion applies. If the interface is multilevel, then the data must be labeled; the Exportation to Multilevel Devices criterion applies. The network architect is free to select an accept- able mechanism for associating a label with an object. With reference to encrypted objects, the following examples are possible: 1. Include a label field in the protocol definition of the object. 2. Implicitly associate the label with the object through the encryption key. That is, the encryption key uniquely identifies a sensitivity level. A sin- gle or private key must be protected at the level of the data that it encrypts. 4.1.1.3.2 Exportation of Labeled Information + Statement from DoD 5200.28-STD The TCB shall designate each communication channel and I/O device as either single-level or multilevel. Any change in this designation shall be done manually and shall be audit- able by the TCB. The TCB shall maintain and be able to audit any change in the sensitivity level or levels associ- ated with a communications channel or I/O device. + Interpretation Each communication channel and network component shall be designated as either single-level or multilevel. Any change in this designation shall be done with the cognizance and approval of the administrator or security officer in charge of the affected components and the administrator or security officer in charge of the NTCB. This change shall be auditable by the network. The NTCB shall maintain and be able to audit any change in the device labels associated with a single-level communication channel or the range asso- ciated with a multilevel communication channel or component. The NTCB shall also be able to audit any change in the set of sensitivity levels associated with the information which can be transmitted over a multilevel communication channel or component. + Rationale Communication channels and components in a network are analogous to communication channels and I/O devices in stand-alone systems. They must be designated as either mul- tilevel (i.e., able to distinguish and maintain separation among information of various sensitivity levels) or single- level. As in the TCSEC, single-level devices may only be attached to single-level channels. The level or set of levels of information that can be sent to a component or over a communication channel shall only change with the knowledge and approval of the security officers (or system administrator, if there is no security officer) of the network, and of the affected components. This requirement ensures that no significant security- relevant changes are made without the approval of all affected parties. 4.1.1.3.2.1 Exportation to Multilevel Devices + Statement from DoD 5200.28-STD When the TCB exports an object to a multilevel I/O device, the sensitivity label associated with that object shall also be exported and shall reside on the same physical medium as the exported information and shall be in the same form (i.e., machine-readable or human-readable form). When the TCB exports or imports an object over a multilevel communi- cations channel, the protocol used on that channel shall provide for the unambiguous pairing between the sensitivity labels and the associated information that is sent or received. + Interpretation The components, including hosts, of a network shall be interconnected over "multilevel communication channels," multiple single-level communication channels, or both, when- ever the information is to be protected at more than a sin- gle sensitivity level. The protocol for associating the sensitivity label and the exported information shall provide the only information needed to correctly associate a sensi- tivity level with the exported information transferred over the multilevel channel between the NTCB partitions in indi- vidual components. This protocol definition must specify the representation and semantics of the sensitivity labels (i.e., the machine-readable label must uniquely represent the sensitivity level). The "unambiguous" association of the sensitivity level with the communicated information shall meet the same level of accuracy as that required for any other label within the NTCB, as specified in the criterion for Label Integrity. This may be provided by protected and highly reliable direct physical layer connections, or by traditional cryptographic link protection in which any errors during transmission can be readily detected, or by use of a separate channel. The range of information imported or exported must be con- strained by the associated device labels. + Rationale This protocol must specify the representation and semantics of the sensitivity labels. See the Mandatory Access Control Policies section in Appendix B. The mul- tilevel device interface to (untrusted) subjects may be implemented either by the interface of the reference moni- tor, per se, or by a multilevel subject (e.g., a "trusted subject" as defined in the Bell-LaPadula Model) that pro- vides the labels based on the internal labels of the NTCB partition. The current state of the art limits the support for mandatory policy that is practical for secure networks. Reference monitor support to ensure the control over all the operations of each subject in the network must be completely provided within the single NTCB partition on which that sub- ject interfaces to the NTCB. This means that the entire portion of the "secure state" represented in the formal security policy model that may be changed by transitions invoked by this subject must be contained in the same component. The secure state of an NTCB partition may be affected by events external to the component in which the NTCB parti- tion resides (e.g., arrival of a message). The effect occurs asynchronusly after being initiated by an event in another component or partition. For example, indeterminate delays may occur between the initiation of a message in one component, the arrival of the message in the NTCB partition in another component, and the corresponding change to the secure state of the second component. Since each component is executing concurrently, to do otherwise would require some sort of network-wide control to synchronize state tran- sitions, such as a global network-wide clock for all proces- sors; in general, such designs are not practical and prob- ably not even desirable. Therefore, the interaction between NTCB partitions is restricted to just communications between pairs (at least logically) of devices-multilevel devices if the device(s) can send/receive data of more than a single level. For broadcast channels the pairs are the sender and intended receiver(s). However, if the broadcast channel carries multiple levels of information, additional mechanism (e.g., cryptochecksum maintained by the TCB) may be required to enforce separation and proper delivery. A common representation for sensitivity labels is needed in the protocol used on that channel and understood by both the sender and receiver when two multilevel devices (in this case, in two different components) are intercon- nected. Each distinct sensitivity level of the overall net- work policy must be represented uniquely in these labels. Within a monolithic TCB, the accuracy of the sensi- tivity labels is generally assured by simple techniques, e.g., very reliable connections over very short physical connections, such as on a single printed circuit board or over an internal bus. In many network environments there is a much higher probability of accidentally or maliciously introduced errors, and these must be protected against. 4.1.1.3.2.2 Exportation to Single-Level Devices + Statement from DoD 5200.28-STD Single-level I/O devices and single-level communication channels are not required to maintain the sensitivity labels of the information they process. However, the TCB shall include a mechanism by which the TCB and an authorized user reliably communicate to designate the single sensitivity level of information imported or exported via single-level communication channels or I/O devices. + Interpretation Whenever one or both of two directly connected com- ponents is not trusted to maintain the separation of information of different sensitivity levels, or whenever the two directly connected components have only a single sensi- tivity level in common, the two components of the network shall communicate over a single-level channel. Single-level components and single-level communication channels are not required to maintain the sensitivity labels of the informa- tion they process. However, the NTCB shall include a reli- able communication mechanism by which the NTCB and an authorized user (via a trusted path) or a subject within an NTCB partition can designate the single sensitivity level of information imported or exported via single-level communica- tion channels or network components. The level of informa- tion communicated must equal the device level. + Rationale Single-level communications channels and single-level components in networks are analogous to single level chan- nels and I/O devices in stand-alone systems in that they are not trusted to maintain the separation of information of different sensitivity levels. The labels associated with data transmitted over those channels and by those components are therefore implicit; the NTCB associates labels with the data because of the channel or component, not because of an explicit part of the bit stream. Note that the sensitivity level of encrypted information is the level of the cipher- text rather than the original level(s) of the plaintext. 4.1.1.3.2.3 Labeling Human-Readable Output + Statement from DoD 5200.28-STD The ADP system administrator shall be able to specify the printable label names associated with exported sensitivity labels. The TCB shall mark the beginning and end of all human-readable, paged, hardcopy output (e.g., line printer output) with human-readable sensitivity labels that prop- erly1 represent the sensitivity of the output. The TCB shall, by default, mark the top and bottom of each page of human-readable, paged, hardcopy output (e.g., line printer output) with human-readable sensitivity labels that prop- erly1 represent the sensitivity of the page. The TCB shall, by default and in an appropriate manner, mark other forms of human readable output (e.g., maps, graphics) with human- readable sensitivity labels that properly1 represent the sensitivity of the output. Any override of these markings defaults shall be auditable by the TCB. _________________________ 1 The hierarchical classification component in human- readable sensitivity labels shall be equal to the greatest hierarchical classification of any of the in- formation in the output that the labels refer to; the non-hierarchical category component shall include all of the non-hierarchical categories of the information in the output the labels refer to, but no other non- hierarchical categories. + Interpretation This criterion imposes no requirement to a component that produces no human-readable output. For those that do produce human-readable output, each sensitivity level that is defined to the network shall have a uniform meaning across all components. The network administrator, in con- junction with any affected component administrator, shall be able to specify the human-readable label that is associated with each defined sensitivity level. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system and partitioned NTCB as defined for these network interpreta- tions. 4.1.1.3.3 Subject Sensitivity Labels + Statement from DoD 5200.28-STD The TCB shall immediately notify a terminal user of each change in the sensitivity level associated with that user during an interactive session. A terminal user shall be able to query the TCB as desired for a display of the subject's complete sensitivity label. + Interpretation An NTCB partition shall immediately notify a terminal user attached to its component of each change in the sensi- tivity level associated with that user. + Rationale The local NTCB partition must ensure that the user understands the sensitivity level of information sent to and from a terminal. When a user has a surrogate process in another component, adjustments to its level may occur to maintain communication with the user. These changes may occur asynchronously. Such adjustments are necessitated by mandatory access control as applied to the objects involved in the communication path. 4.1.1.3.4 Device Labels + Statement from DoD 5200.28-STD The TCB shall support the assignment of minimum and maximum sensitivity levels to all attached physical devices. These sensitivity levels shall be used by the TCB to enforce con- straints imposed by the physical environments in which the devices are located. + Interpretation This requirement applies as written to each NTCB parti- tion that is trusted to separate information based on sensi- tivity level. Each I/O device in a component, used for com- munication with other network components, is assigned a dev- ice range, consisting of a set of labels with a maximum and minimum. (A device range usually contains, but does not necessarily contain, all possible labels "between" the max- imum and minimum, in the sense of dominating the minimum and being dominated by the maximum.) The NTCB always provides an accurate label for informa- tion exported through devices. Information exported or imported using a single-level device is labelled implicitly by the sensitivity level of the device. Information exported from one multilevel device and imported at another must be labelled through an agreed-upon protocol, unless it is labelled implicitly by using a communication link that always carries a single level. Information exported at a given sensitivity level can be sent only to an importing device whose device range con- tains that level or a higher level. If the importing device range does not contain the given level, the information is relabelled upon reception at a higher level within the importing device range. Relabelling should not occur other- wise. + Rationale The purpose of device labels is to reflect and con- strain the sensitivity levels of information authorized for the physical environment in which the devices are located. The information transfer restrictions permit one-way communication (i.e., no acknowledgements) from one device to another whose ranges have no level in common, as long as each level in the sending device range is dominated by some level in the receiving device range. It is never permitted to send information at a given level to a device whose range does not contain a dominating level. (See Appendix C for similar interconnection rules for the interconnected AIS view.) 4.1.1.4 Mandatory Access Control + Statement from DoD 5200.28-STD The TCB shall enforce a mandatory access control policy over all resources (i.e., subjects, storage objects, and I/O dev- ices) that are directly or indirectly accessible by subjects external to the TCB. These subjects and objects shall be assigned sensitivity labels that are a combination of hierarchical classification levels and non-hierarchical categories, and the labels shall be used as the basis for mandatory access control decisions. The TCB shall be able to support two or more such sensitivity levels. (See the Mandatory Access Control interpretations.) The following requirements shall hold for all accesses between all sub- jects external to the TCB and all objects directly or indirectly accessible by these subjects. A subject can read an object only if the hierarchical classification in the subject's sensitivity level is greater than or equal to the hierarchical classification of the object's sensitivity level and the non-hierarchical categories in the subject's sensitivity level include all the non-hierarchical categories in the object's sensitivity level. A subject can write an object only if the hierarchical classification in the subject's sensitivity level is less than or equal to the hierarchical classification of the object's sensitivity level and the non-hierarchical categories in the subject's sensitivity level are included in the non-hierarchical categories in the object's sensitivity level. Identification and authentication data shall be used by the TCB to authen- ticate the user's identity and to ensure that the sensi- tivity level and authorization of subjects external to the TCB that may be created to act on behalf of the individual user are dominated by the clearance and authorization of that user. + Interpretation Each partition of the NTCB exercises mandatory access control policy over all subjects and objects in its com- ponent. In a network, the responsibility of an NTCB parti- tion encompasses all mandatory access control functions in its component that would be required of a TCB in a stand- alone system. In particular, subjects and objects used for communication with other components are under the control of the NTCB partition. Mandatory access control includes secrecy and integrity control to the extent that the network sponsor has described in the overall network security pol- icy. Conceptual entities associated with communication between two components, such as sessions, connections and virtual circuits, may be thought of as having two ends, one in each component, where each end is represented by a local object. Communication is viewed as an operation that copies information from an object at one end of a communication path to an object at the other end. Transient data-carrying entities, such as datagrams and packets, exist either as information within other objects, or as a pair of objects, one at each end of the communication path. The requirement for "two or more" sensitivity levels can be met by either secrecy or integrity levels. When there is a mandatory integrity policy, the stated require- ments for reading and writing are generalized to: A subject can read an object only if the subject's sensitivity level dominates the object's sensitivity level, and a subject can write an object only if the object's sensitivity level dom- inates the subject's sensitivity level. Based on the integrity policy, the network sponsor shall define the domi- nance relation for the total label, for example, by combin- ing secrecy and integrity lattices. - + Rationale An NTCB partition can maintain access control only over subjects and objects in its component. At levels B2 and above, the NTCB partition must maintain access control over all subjects and objects in its component. Access by a sub- ject in one component to information contained in an object in another component requires the creation of a subject in the remote component which acts as a surrogate for the first subject. The mandatory access controls must be enforced at the interface of the reference monitor (viz. the mechanism that controls physical processing resources) for each NTCB parti- tion. This mechanism creates the abstraction of subjects and objects which it controls. Some of these subjects out- side the reference monitor, per se, may be designated to implement part of an NTCB partition's mandatory policy, e.g., by using the ``trusted subjects" defined in the Bell- LaPadula model. The prior requirements on exportation of labeled infor- mation to and from I/O devices ensure the consistency between the sensitivity labels of objects connected by a communication path. As noted in the introduction, the net- work architecture must recognize the linkage between the overall mandatory network security policy and the connection oriented abstraction. For example, individual data-carrying entities such as datagrams can have individual sensitivity labels that subject them to mandatory access control in each component. The abstraction of a single-level connection is realized and enforced implicitly by an architecture while a connection is realized by single-level subjects that neces- sarily employ only datagrams of the same level. The fundamental trusted systems technology permits the DAC mechanism to be distributed, in contrast to the require- ments for mandatory access control. For networks this separation of MAC and DAC mechanisms is the rule rather than _________________________ - See, for example, Grohn, M. J., A Model of a Pro- _ _____ __ _ ___ tected Data Management System, ESD-TR-76-289, I. P. ______ ____ __________ ______ Sharp Assoc. Ltd., June, 1976; and Denning, D .E., Lunt, T. F., Neumann, P. G., Schell, R. R., Heckman, M. and Shockley, W., Secure Distributed Data Views, Secu- ______ ___________ ____ _____ ____ rity Policy and Interpretation for a Class A1 Multilev- ____ ______ ___ ______________ ___ _ _____ __ ________ el Secure Relational Database System,SRI International, __ ______ __________ ________ ______ November 1986. the exception. The set of total sensitivity labels used to represent all the sensitivity levels for the mandatory access control (combined data secrecy and data integrity) policy always forms a partially ordered set. Without loss of generality, this set of labels can always be extended to form a lattice, by including all the combinations of non-hierarchical categories. As for any lattice, a dominance relation is always defined for the total sensitivity labels. For admin- istrative reasons it may be helpful to have a maximum level which dominates all others. 4.1.2 Accountability _ _ _ ______________ 4.1.2.1 Identification and Authentication + Statement from DoD 5200.28-STD The TCB shall require users to identify themselves to it before beginning to perform any other actions that the TCB is expected to mediate. Furthermore, the TCB shall maintain authentication data that includes information for verifying the identify of individual users (e.g., passwords) as well as information for determining the clearance and authoriza- tions of individual users. This data shall be used by the TCB to authenticate the user's identity and to ensure that the sensitivity level and authorization of subjects external to the TCB that may be created to act on behalf of the indi- vidual user are dominated by the clearance and authorization of that user. The TCB shall protect authentication data so that it cannot be accessed by any unauthorized user. The TCB shall be able to enforce individual accountability by providing the capability to uniquely identify each indivi- dual ADP system user. The TCB shall also provide the capa- bility of associating this identity with all auditable actions taken by that individual. + Interpretation The requirement for identification and authentication of users is the same for a network system as for an ADP sys- tem. The identification and authentication may be done by the component to which the user is directly connected or some other component, such as an identification and authen- tication server. Available techniques, such as those described in the Password Guideline=, are generally also applicable in the network context. However, in cases where the NTCB is expected to mediate actions of a host (or other network component) that is acting on behalf of a user or group of users, the NTCB may employ identification and _________________________ = Department of Defense Password Management Guide- __________ __ _______ ________ __________ _____ line, CSC-STD-002-85 ____ authentication of the host (or other component) in lieu of identification and authentication of an individual user, so long as the component identifier implies a list of specific users uniquely associated with the identifier at the time of its use for authentication. This requirement does not apply to internal subjects. Authentication information, including the identity of a user (once authenticated) may be passed from one component to another without reauthentication, so long as the NTCB protects (e.g., by encryption) the information from unau- thorized disclosure and modification. This protection shall provide at least a similar level of assurance (or strength of mechanism) as pertains to the protection of the authenti- cation mechanism and authentication data. + Rationale The need for accountability is not changed in the con- text of a network system. The fact that the NTCB is parti- tioned over a set of components neither reduces the need nor imposes new requirements. That is, individual accountabil- ity is still the objective. Also, in the context of a net- work system at the (C2) level or higher "individual accoun- tability" can be satisfied by identification of a host (or other component) so long as the requirement for traceability to individual users or a set of specific individual users with active subjects is satisfied. There is allowed to be an uncertainty in traceability because of elapsed time between changes in the group membership and the enforcement in the access control mechanisms. In addition, there is no need in a distributed processing system like a network to reauthen- ticate a user at each point in the network where a projec- tion of a user (via the subject operating on behalf of the user) into another remote subject takes place. The passing of identifiers and/or authentication infor- mation from one component to another is usually done in sup- port to the implementation of the discretionary access con- trol (DAC). This support relates directly to the DAC regarding access by a user to a storage object in a dif- ferent NTCB partition than the one where the user was authenticated. Employing a forwarded identification implies additional reliance on the source and components along the path. If the authenticated identification is used as the basis of determining a sensitivity label for a subject, it must satisfy the Label Integrity criterion. An authenticated identification may be forwarded between components and employed in some component to iden- tify the sensitivity level associated with a subject created to act on behalf of the user so identified. 4.1.2.1.1 Trusted Path + Statement from DoD 5200.28-STD The TCB shall support a trusted communication path between itself and users for use when a positive TCB-to-user connec- tion is required (e.g., login, change subject sensitivity level). Communications via this trusted path shall be activated exclusively by a user or the TCB and shall be logically and unmistakably distinguishable from other paths. + Interpretation A trusted path is supported between a user (i.e., human) and the NTCB partition in the component to which the user is directly connected. + Rationale When a user logs into a remote component, the user id is transmitted securely between the local and remote NTCB partitions in accordance with the requirements in Identifi- cation and Authentication. Trusted Path is necessary in order to assure that the user is communicating with the NTCB and only the NTCB when security relevant activities are taking place (e.g., authen- ticate user, set current session sensitivity level). How- ever, Trusted Path does not address communications within the NTCB, only communications between the user and the NTCB. If, therefore, a component does not support any direct user communication then the component need not contain mechanisms for assuring direct NTCB to user communications. The requirement for trusted communication between one NTCB partition and another NCTB partition is addressed in the System Architecture section. These requirements are separate and distinct from the user to NTCB communication requirement of a trusted path. However, it is expected that this trusted communication between one NTCB partition and another NTCB partition will be used in conjunction with the trusted path to implement trusted communication between the user and the remote NTCB partition. 4.1.2.2 Audit + Statement from DoD 5200.28-STD The TCB shall be able to create, maintain, and protect from modification or unauthorized access or destruction an audit trail of accesses to the objects it protects. The audit data shall be protected by the TCB so that read access to it is limited to those who are authorized for audit data. The TCB shall be able to record the following types of events: use of identification and authentication mechanisms, introduction of objects into a user's address space (e.g., file open, program initiation), deletion of objects, actions taken by computer operators and system administrators and/or system security officers, and other security relevant events. The TCB shall also be able to audit any override of human-readable output markings. For each recorded event, the audit record shall identify: date and time of the event, user, type of event, and success or failure of the event. For identification/authentication events the origin of request (e.g., terminal ID) shall be included in the audit record. For events that introduce an object into a user's address space and for object deletion events the audit record shall include the name of the object and the object's sensitivity level. The ADP system administrator shall be able to selectively audit the actions of any one or more users based on individual identify and/or object sensitivity level. The TCB shall be able to audit the identified events that may be used in the exploitation of covert storage channels. The TCB shall contain a mechanism that is able to monitor the occurrence or accumulation of secu- rity auditable events that may indicate an imminent viola- tion of security policy. This mechanism shall be able to immediately notify the security administrator when thres- holds are exceeded and, if the occurrence or accumulation of these security relevant events continues, the system shall take the least disruptive action to terminate the event. + Interpretation This criterion applies as stated. The sponsor must select which events are auditable. If any such events are not distinguishable by the NTCB alone (for example those identified in Part II), the audit mechanism shall provide an interface, which an authorized subject can invoke with parameters sufficient to produce an audit record. These audit records shall be distinguishable from those provided by the NTCB. In the context of a network system, "other security relevant events" (depending on network system architecture and network security policy) might be as fol- lows: 1. Identification of each access event (e.g., estab- lishing a connection or a connectionless association between processes in two hosts of the network) and its principal parameters (e.g., host identifiers of the two hosts involved in the access event and user identifier or host identifier of the user or host that is requesting the access event) 2. Identification of the starting and ending times of each access event using local time or global syn- chronized time 3. Identification of security-relevant exceptional con- ditions (e.g., potential violation of data integrity, such as misrouted datagrams) detected during the transactions between two hosts 4. Utilization of cryptographic variables 5. Changing the configuration of the network (e.g., a component leaving the network and rejoining) In addition, identification information should be included in appropriate audit trail records, as necessary, to allow association of all related (e.g., involving the same network event) audit trail records (e.g., at different hosts) with each other. Furthermore, a component of the network system may provide the required audit capability (e.g., storage, retrieval, reduction, analysis) for other components that do not internally store audit data but transmit the audit data to some designated collection com- ponent. Provisions shall be made to control the loss of audit data due to unavailability of resources. In the context of a network system, the "user's address space" is extended, for object introduction and deletion events, to include address spaces being employed on behalf of a remote user (or host). However, the focus remains on users in contrast to internal subjects as discussed in the DAC criterion. In addition, audit information must be stored in machine-readable form. The capability must exist to audit the identified events that may be used in the exploitation of covert storage channels. To accomplish this, each NTCB partition must be able to audit those events locally that may lead to the exploitation of a covert storage channel which exist because of the network. The sponsor shall identify the specific auditable events that may indicate an imminent violation of security policy. The component which detects the occurrence or accu- mulation of such events must be able to notify an appropri- ate administrator when thresholds are exceeded, and to ini- tiate actions which will result in termination of the event if the accumulation continues. For example, when the thres- hold of unsuccessful login attempts within a period of time is exceeded, login shall be inhibited for a specific time period. + Rationale For remote users, the network identifiers (e.g., inter- net address) can be used as identifiers of groups of indivi- dual users (e.g., "all users at Host A") to eliminate the maintenance that would be required if individual identifica- tion of remote users was employed. In this class (C2), how- ever, it must be possible to identify (immediately or at some later time) the individuals represented by a group identifier. In all other respects, the interpretation is a straightforward extension of the criterion into the context of a network system. Identification of covert channel events is addressed in the Covert Channel Analysis section. Because of concurrency and synchronization problems, it may not be possible to detect in real time the accumulation of security auditable events that are occurring in different NTCB partitions. However, each NTCB partition that has been allocated audit responsibility must have the capability to detect the local accumulation of events, to notify the par- tition security administrator and/or the network security administrator, and to initiate actions which will result in termination of the event locally. 4.1.3 Assurance _ _ _ _________ 4.1.3.1 Operational Assurance 4.1.3.1.1 System Architecture + Statement from DoD 5200.28-STD The TCB shall maintain a domain for its own execution that protects it from external interference or tampering (e.g., by modification of its code or data structures). The TCB shall maintain process isolation through the provision of distinct address spaces under its control. The TCB shall be internally structured into well-defined largely independent modules. It shall make effective use of available hardware to separate those elements that are protection-critical from those that are not. The TCB modules shall be designed such that the principle of least privilege is enforced. Features in hardware, such as segmentation, shall be used to support logically distinct storage objects with separate attributes (namely: readable, writable). The user interface to the TCB shall be completely defined and all elements of the TCB identified. The TCB shall be designed and structured to use a complete, conceptually simple protection mechanism with precisely defined semantics. This mechanism shall play a central role in enforcing the internal structuring of the TCB and the system. The TCB shall incorporate significant use of layering, abstraction and data hiding. Significant system engineering shall be directed toward minimizing the complexity of the TCB and excluding from the TCB modules that are not protection-critical. + Interpretation The system architecture criterion must be met individu- ally by all NTCB partitions. Implementation of the require- ment that the NTCB maintain a domain for its own execution is achieved by having each NTCB partition maintain a domain for its own execution. Since each component is itself a dis- tinct domain in the overall network system, this also satis- fies the requirement for process isolation through distinct address spaces in the special case where a component has only a single subject. The NTCB must be internally structured into well- defined largely independent modules and meet the hardware requirements. This is satisfied by having each NTCB parti- tion so structured. The NTCB controls all network resources. These resources are the union of the sets of resources over which the NTCB partitions have control. Code and data structures belonging to the NTCB, transferred among NTCB subjects (i.e., subjects outside the reference monitor but inside the NTCB) belonging to different NTCB partitions, must be protected against external interference or tamper- ing. For example, a cryptographic checksum or physical means may be employed to protect user authentication data exchanged between NTCB partitions. Each NTCB partition must enforce the principle of least privilege within its component. Additionally, the NTCB must be structured so that the principle of least privilege is enforced in the system as a whole. The NTCB must be designed and structured according to the network security architecture to use a complete, concep- tually simple protection mechanism. Furthermore, each NTCB partition must also be so designed and structured. Significant system engineering should be directed toward minimizing the complexity of each NTCB partition, and of the NTCB. Care shall be taken to exclude modules (and components) that are not protection-critical from the NTCB. It is recognized that some modules and/or components may need to be included in the NTCB and must meet the NTCB requirements even though they may not appear to be directly protection-critical. The correct operation of these modules/components is necessary for the correct operation of the protection-critical modules and components. However, the number and size of these modules/components should be kept to a minimum. Each NTCB partition provides isolation of resources (within its component) in accord with the network system architecture and security policy so that "supporting ele- ments" (e.g., DAC and user identification) for the security mechanisms of the network system are strengthened compared to C2, from an assurance point of view, through the provi- sion of distinct address spaces under control of the NTCB. As discussed in the Discretionary Access Control sec- tion, the DAC mechanism of a NTCB partition may be imple- mented at the interface of the reference monitor or may be distributed in subjects that are part of the NTCB in the same or different component. When distributed in NTCB sub- jects (i.e., when outside the reference monitor), the assurance requirements for the design and implementation of the DAC shall be those of class C2 for all networks of class C2 or above. + Rationale The requirement that the NTCB be structured into modules and meet the hardware requirements applies within the NTCB partitions in the various components. The principle of least privilege requires that each user or other individual with access to the system be given only those resources and authorizations required for the performance of this job. In order to enfore this principle in the system it must be enforced in every NTCB partition that supports users or other individuals. For example, prohibiting access by administrators to objects outside the NTCB partition (e.g., games) lessens the opportunity of dam- age by a Trojan Horse. The requirement for the protection of communications between NTCB partitions is specifically directed to subjects that are part of the NTCB partitions. Any requirements for such protection for the subjects that are outside the NTCB partitions are addressed in response to the integrity requirements of the security policy. There are certain parts of a network (modules and/or components) that may not appear to be directly protection- critical in that they are not involved in access control decisions, do not directly audit, and are not involved in the identification/authentication process. However, the security of the network must depend on the correct operation of these modules and/or components. An example of this is a single level packet switch. Although it may not normally be involved directly in enforcing the discretionary security policy, this switch may be trusted not to mix data from dif- ferent message streams. If the switch does not operate correctly, data could get mixed, and unauthorized access could result. Therefore, these modules/components must be included in the NTCB and must meet the NTCB requirements applicable to the policy element(s) for which they are responsible. 4.1.3.1.2 System Integrity + Statement from DoD 5200.28-STD Hardware and/or software features shall be provided that can be used to periodically validate the correct operation of the on-site hardware and firmware elements of the TCB. + Interpretation Implementation of the requirement is partly achieved by having hardware and/or software features that can be used to periodically validate the correct operation of the hardware and firmware elements of each component's NTCB partition. Features shall also be provided to validate the identity and correct operation of a component prior to its incorporation in the network system and throughout system operation. For example, a protocol could be designed that enables the com- ponents of the partitioned NTCB to exchange messages period- ically and validate each other's correct response. The pro- tocol shall be able to determine the remote entity's ability to respond. NTCB partitions shall provide the capability to report to network administrative personnel the failures detected in other NTCB partitions. Intercomponent protocols implemented within a NTCB shall be designed in such a way as to provide correct opera- tion in the case of failures of network communications or individual components. The allocation of mandatory and dis- cretionary access control policy in a network may require communication between trusted subjects that are part of the NTCB partitions in different components. This communication is normally implemented with a protocol between the subjects as peer entities. Incorrect access within a component shall not result from failure of an NTCB partition to communicate with other components. + Rationale The first paragraph of the interpretation is a straightforward extension of the requirement into the con- text of a network system and partitioned NTCB as defined for these network criteria. NTCB protocols should be robust enough so that they permit the system to operate correctly in the case of local- ized failure. The purpose of this protection is to preserve the integrity of the NTCB itself. It is not unusual for one or more components in a network to be inoperative at any time, so it is important to minimize the effects of such failures on the rest of the network. Additional integrity and denial of service issues are addressed in Part II. It should be clear that some integrity and denial of service features can reside outside the NTCB. Otherwise all software in a network would be in the NTCB. Every piece of software that has an opportunity to write to some data or protocol field is "trusted" to preserve integrity or not cause denial of service to some extent. For example, it is necessary to "trust" TELNET to correctly translate user data, and to eventually transmit packets. FTP also has to be "trusted" to not inappropriately modify files, and to attempt to complete the file transfer. These protocols can be designed, however to exist outside the NTCB (from a pro- tection perspective). It is beneficial to do this type of security engineering so that the amount of code that must be trusted to not disclose data is minimized. Putting every- thing inside the NTCB contradicts the requirement to perform "significant system engineering ... directed toward ... excluding from the TCB modules that are not protection crit- ical," which removes the primary difference between B2 and B3. If everything has to be in the TCB to ensure data integrity and protection against denial of service, there will be considerably less assurance that disclosure protec- tion is maximized. 4.1.3.1.3 Covert Channel Analysis + Statement from DoD 5200.28-STD The system developer shall conduct a thorough search for covert channels and make a determination (either by actual measurement or by engineering estimation) of the maximum bandwidth of each identified channel. (See the Covert Chan- nels Guideline section.) FORMAL METHODS SHALL BE USED IN THE ANALYSIS. + Interpretation The requirement, including the TCSEC Covert Channel Guideline, applies as written. In a network, there are additional instances of covert channels associated with com- munication between components. THE FORMAL METHODS SHALL BE USED IN THE ANALYSIS OF EACH INDIVIDUAL COMPONENT DESIGN AND IMPLEMENTATION. + Rationale The exploitation of network protocol information (e.g., headers) can result in covert storage channels. Exploitation of frequency of transmission can result in covert timing channels. The topic has been addressed in the literature.- 4.1.3.1.4 Trusted Facility Management + Statement from DoD 5200.28-STD The TCB shall support separate operator and administrator functions. The functions performed in the role of a secu- rity administrator shall be identified. The ADP system administrative personnel shall only be able to perform secu- rity administrator functions after taking a distinct audit- able action to assume the security administrator role on the ADP system. Non-security functions that can be performed in the security administration role shall be limited strictly _________________________ - See, for example, Girling, C. G., "Covert Channels in LAN's," IEEE Transactions on Software Engineering, ____ ____________ __ ________ ___________ Vol. SE-13, No. 2, February 1987; and Padlipsky, M. A., Snow, D. P., and Karger, P. A., Limitations of End-to- ___________ __ ___ __ End Encryption in Secure Computer Networks, MITRE ___ __________ __ ______ ________ ________ Technical Report, MTR-3592, Vol. I, May 1978 (ESD TR 78-158, DTIC AD A059221). to those essential to performing the security role effec- tively. + Interpretation This requirement applies as written to both the network as a whole and to individual components which support such personnel. + Rationale It is recognized that based on the allocated policy elements some components may operate with no human inter- face. 4.1.3.1.5 Trusted Recovery + Statement from DoD 5200.28-STD Procedures and/or mechanisms shall be provided to assure that, after an ADP system failure or other discontinuity, recovery without a protection compromise is obtained. + Interpretation The recovery process must be accomplished without a protection compromise after the failure or other discon- tinuity of any NTCB partition. It must also be accomplished after a failure of the entire NTCB. + Rationale This is a straight-forward extension of the requirement into the network context, and takes into account that it is possible for parts of the system to fail while other parts continue to operate normally. This may be a security- relevant event; if so it must be audited. 4.1.3.2 Life-Cycle Assurance 4.1.3.2.1 Security Testing + Statement from DoD 5200.28-STD The security mechanisms of the ADP system shall be tested and found to work as claimed in the system documentation. A team of individuals who thoroughly understand the specific implementation of the TCB shall subject its design documen- tation, source code, and object code to through analysis and testing. Their objectives shall be: to uncover all design and implementation flaws that would permit a subject exter- nal to the TCB to read, change, or delete data normally denied under the mandatory or discretionary security policy enforced by the TCB; as well as to assure that no subject (without authorization to do so) is able to cause the TCB to enter a state such that it is unable to respond to communi- cations initiated by other users. The TCB shall be found resistant to penetration. All discovered flaws shall be removed or neutralized and the TCB retested to demonstrate that they have been eliminated and that new flaws have not been introduced. Testing shall demonstrate that the TCB implementation is consistent with the descriptive top-level specification. No design flaws and no more than a few correctable implementation flaws may be found during testing and there shall be reasonable confidence that few remain. MANUAL OR OTHER MAPPING OF THE FTLS TO THE SOURCE CODE MAY FORM A BASIS FOR PENETRATION TESTING. (See the Security Testing Guidelines.) + Interpretation Testing of a component will require a testbed that exercises the interfaces and protocols of the component including tests under exceptional conditions. The testing of a security mechanism of the network system for meeting this criterion shall be an integrated testing procedure involving all components containing an NTCB partition that implement the given mechanism. This integrated testing is additional to any individual component tests involved in the evaluation of the network system. The sponsor should iden- tify the allowable set of configurations including the sizes of the networks. Analysis or testing procedures and tools shall be available to test the limits of these configura- tions. A change in configuration within the allowable set of configurations does not require retesting. The testing of each component will include the intro- duction of subjects external to the NTCB partition for the component that will attempt to read, change, or delete data normally denied. If the normal interface to the component does not provide a means to create the subjects needed to conduct such a test, then this portion of the testing shall use a special version of the untrusted software for the com- ponent that results in subjects that make such attempts. The results shall be saved for test analysis. Such special versions shall have an NTCB partition that is identical to that for the normal configuration of the component under evaluation. The testing of the mandatory controls shall include tests to demonstrate that the labels for information imported and/or exported to/from the component accurately represent the labels maintained by the NTCB partition for the component for use as the basis for its mandatory access control decisions. The tests shall include each type of device, whether single-level or multilevel, supported by the component. The NTCB must be found resistant to penetration. This applies to the NTCB as a whole, and to each NTCB partition in a component of this class. + Rationale The phrase "no subject (without authorization to do so) is able to cause the TCB to enter a state such that it is unable to respond to communications initiated by other users" relates to the security services (Part II of this TNI) for the Denial of Service problem, and to correctness of the protocol implementations. Testing is an important method available in this evaluation division to gain any assurance that the security mechanisms perform their intended function. A major purpose of testing is to demonstrate the system's response to inputs to the NTCB partition from untrusted (and possibly mali- cious) subjects. In contrast to general purpose systems that allow for the dynamic creation of new programs and the introductions of new processes (and hence new subjects) with user speci- fied security properities, many network components have no method for introducing new programs and/or processes during their normal operation. Therefore, the programs necessary for the testing must be introduced as special versions of the software rather than as the result of normal inputs by the test team. However, it must be insured that the NTCB partition used for such tests is identical to the one under evaluation. Sensitivity labels serve a critical role in maintaining the security of the mandatory access controls in the net- work. Especially important to network security is the role of the labels for information communicated between com- ponents - explicit labels for multilevel devices and impli- cit labels for single-level devices. Therefore the testing for correct labels is highlighted. The requirement for testing to demonstrate consistency between the NTCB implementation and the FTLS is a straight- forward extension of the TCSEC requirement into the context of a network system. 4.1.3.2.2 Design Specification and Verification + Statement from DoD 5200.28-STD A formal model of the security policy supported by the TCB shall be maintained over the life cycle of the ADP system that is proven and demonstrated to be consistent with its axioms. A descriptive top-level specification (DTLS) of the TCB shall be maintained that completely and accurately describes the TCB in terms of exceptions, error messages, and effects. A FORMAL TOP-LEVEL SPECIFICATION (FTLS) OF THE TCB SHALL BE MAINTAINED THAT ACCURATELY DESCRIBES THE TCB IN TERMS OF EXCEPTIONS, ERROR MESSAGES, AND EFFECTS. THE DTLS AND FTLS SHALL INCLUDE THOSE COMPONENTS OF THE TCB THAT ARE IMPLEMENTED AS HARDWARE AND/OR FIRMWARE IF THEIR PROPERTIES ARE VISIBLE AT THE TCB INTERFACE. THE FTLS SHALL BE SHOWN to be an accurate description of the TCB interface. A con- vincing argument shall be given that the DTLS is consistent with the model AND A COMBINATION OF FORMAL AND INFORMAL TECHNIQUES SHALL BE USED TO SHOW THAT THE FTLS IS CONSISTENT WITH THE MODEL. THIS VERIFICATION EVIDENCE SHALL BE CON- SISTENT WITH THAT PROVIDED WITHIN THE STATE-OF-THE-ART OF THE PARTICULAR NATIONAL COMPUTER SECURITY CENTER-ENDORSED FORMAL SPECIFICATION AND VERIFICATION SYSTEM USED. MANUAL OR OTHER MAPPING OF THE FTLS TO THE TCB SOURCE CODE SHALL BE PERFORMED TO PROVIDE EVIDENCE OF CORRECT IMPLEMENTATION. + Interpretation The overall network security policy expressed in this model will provide the basis for the mandatory access con- trol policy exercised by the NTCB over subjects and storage objects in the entire network. The policy will also be the basis for the discretionary access control policy exercised by the NTCB to control access of named users to named objects. Data integrity requirements addressing the effects of unauthorized MSM need not be included in this model. The overall network policy must be decomposed into policy ele- ments that are allocated to appropriate components and used as the basis for the security policy model for those com- ponents. The level of abstraction of the model, and the set of subjects and objects that are explicitly represented in the model, will be affected by the NTCB partitioning. Subjects and objects must be represented explicitly in the model for the partition if there is some network component whose NTCB partition exercises access control over them. The model shall be structured so that the axioms and entities applica- ble to individual network components are manifest. Global network policy elements that are allocated to components shall be represented by the model for that component. AN FTLS FOR A NETWORK CONSISTS OF A COMPONENT FTLS FOR EACH UNIQUE TRUSTED NETWORK COMPONENT, PLUS ANY GLOBAL DECLARATIONS AND ASSERTIONS THAT APPLY TO MORE THAN ONE COM- PONENT. IF THE MODEL FOR EACH COMPONENT REPRESENTS ALL THE GLOBAL MANDATORY POLICY ELEMENTS ALLOCATED TO THAT COM- PONENT, THERE MAY NOT BE ANY GLOBAL ASSERTIONS NEEDED, AND IN THIS CASE THE COLLECTION OF COMPONENT FTLS, WITH ANY SHARED DECLARATIONS, IS THE NETWORK FTLS. EACH COMPONENT FTLS SHALL DESCRIBE THE INTERFACE TO THE NTCB PARTITION OF ITS COMPONENTS. The requirements for a network DTLS are given in the Design Documentation section. + Rationale The treatment of the model depends to a great extent on the degree of integration of the communications service into a distributed system. In a closely coupled distributed sys- tem, one might use a model that closely resembles one appropriate for a stand-alone computer system. In all cases, the model of each partition will be expected to show the role of the NTCB partition in each kind of component. It will most likely clarify the model, although not part of the model, to show access restrictions implied by the system design; for example, subjects representing protocol entities might have access only to objects containing data units at the same layer of protocol. The allocation of subjects and objects to different proto- col layers is a protocol design choice which need not be reflected in the security policy model. THE FTLS MUST REPRESENT THE UNDERLYING REFERENCE MONI- TOR AND ANY SUBJECTS IMPLEMENTING THE MANDATORY POLICY. OTHER POLICY ELEMENTS DISTRIBUTED IN NTCB SUBJECTS (SEE THE INTERPRETATION OF SYSTEM ARCHITECTURE) NEED NOT BE REPRESENTED BY THE FTLS. 4.1.3.2.3 Configuration Management + Statement from DoD 5200.28-STD During THE ENTIRE LIFE-CYCLE, I.E. DURING THE DESIGN, DEVELOPMENT, and maintenance of the TCB, a configuration management system shall be in place FOR ALL SECURITY- RELEVANT HARDWARE, FIRMWARE, AND SOFTWARE that maintains control of changes to THE FORMAL MODEL, the descriptive AND FORMAL top-level SPECIFICATIONS, other design data, imple- mentation documentation, source code, the running version of the object code, and test fixtures and documentation. The configuration management system shall assure a consistent mapping among all documentation and code associated with the current version of the TCB. Tools shall be provided for generation of a new version of the TCB from source code. Also available shall be tools, MAINTAINED UNDER STRICT CON- FIGURATION CONTROL, for comparing a newly generated version with the previous TCB version in order to ascertain that only the intended changes have been made in the code that will actually be used as the new version of the TCB. A COM- BINATION OF TECHNICAL, PHYSICAL, AND PROCEDURAL SAFEGUARDS SHALL BE USED TO PROTECT FROM UNAUTHORIZED MODIFICATION OR DESTRUCTION THE MASTER COPY OR COPIES OF ALL MATERIAL USED TO GENERATE THE TCB. + Interpretation The requirement applies as written, with the following extensions: 1. A configuration management system must be in place for each NTCB partition. 2. A configuration management plan must exist for the entire system. If the configuration management sys- tem is made up of the conglomeration of the confi- guration management systems of the various NTCB par- titions, then the configuration management plan must address the issue of how configuration control is applied to the system as a whole. ALL MATERIAL USED IN GENERATING A NEW VERSION OF THE NTCB AND EACH NTCB PARTITION MUST BE PROTECTED, REGARDLESS OF WHERE IT PHYSICALLY RESIDES. + Rationale Each NTCB partition must have a configuration manage- ment system in place, or else there will be no way for the NTCB as a whole to have an effective configuration manage- ment system. The other extensions are merely reflections of the way that networks operate in practice. THIS NEW REQUIREMENT EXPLICITLY MANDATES THE PROTECTION OF MATERIAL USED TO GENERATE AN NTCB PARTITION, EVEN WHEN THE GENERATION OCCURS BY DOWN-LINE LOADING OF A REMOTE COM- PONENT. 4.1.3.2.4 Trusted Distribution + Statement from DoD 5200.28-STD A TRUSTED ADP SYSTEM CONTROL AND DISTRIBUTION FACILITY SHALL BE PROVIDED FOR MAINTAINING THE INTEGRITY OF THE MAPPING BETWEEN THE MASTER DATA DESCRIBING THE CURRENT VERSION OF THE TCB AND THE ON-SITE MASTER COPY OF THE CODE FOR THE CURRENT VERSION. PROCEDURES (E.G., SITE SECURITY ACCEPTANCE TESTING) SHALL EXIST FOR ASSURING THAT THE TCB SOFTWARE, FIRMWARE, AND HARDWARE UPDATES DISTRIBUTED TO A CUSTOMER ARE EXACTLY AS SPECIFIED BY THE MASTER COPIES. + Interpretation THIS REQUIREMENT APPLIES AS STATED, WITH THE ADDITIONAL REQUIREMENT THAT, IF DOWN-LINE LOADING IS USED, THERE MUST BE A TRUSTED METHOD OF GENERATING, SENDING, AND LOADING ANY SOFTWARE INVOLVED. + Rationale THIS IS A STRAIGHTFORWARD EXTENSION OF THE REQUIREMENT INTO THE NETWORK CONTEXT. 4.1.4 Documentation. _ _ _ _____________ 4.1.4.1 Security Features User's Guide + Statement from DoD 5200.28-STD A single summary, chapter, or manual in user documentation shall describe the protection mechanisms provided by the TCB, interpretations on their use, and how they interact with one another. + Interpretation This user documentation describes user visible protec- tion mechanisms at the global (network system) level and at the user interface of each component, and the interaction among these. + Rationale The interpretation is an extension of the requirement into the context of a network system as defined for these network criteria. Documentation of protection mechanisms provided by individual components is required by the cri- teria for trusted computer systems that are applied as appropriate for the individual components. 4.1.4.2 Trusted Facility Manual + Statement from DoD 5200.28-STD A manual addressed to the ADP system administrator shall present cautions about functions and privileges that should be controlled when running a secure facility. The procedures for examining and maintaining the audit files as well as the detailed audit record structure for each type of audit event shall be given. The manual shall describe the operator and administrator functions related to security, to include changing the security characteristics of a user. It shall provide interpretations on the consistent and effective use of the protection features of the system, how they interact, how to securely generate a new TCB, and facility procedures, warnings, and privileges that need to be controlled in order to operate the facility in a secure manner. The TCB modules that contain the reference validation mechanism shall be identified. The procedures for secure generation of a new TCB from source after modification of any modules in the TCB shall be described. It shall include the procedures to ensure that the system is initially started in a secure manner. Procedures shall also be included to resume secure system operation after any lapse in system operation. + Interpretation This manual shall contain specifications and procedures to assist the system administrator(s) maintain cognizance of the network configuration. These specifications and pro- cedures shall address the following: 1. The hardware configuration of the network itself; 2. The implications of attaching new components to the network; 3. The case where certain components may periodically leave the network (e.g., by crashing, or by being disconnected) and then rejoin; 4. Network configuration aspects that can impact the security of the network system; (For example, the manual should describe for the network system administrator the interconnections among components that are consistent with the overall network system architecture.) 5. Loading or modifying NTCB software or firmware (e.g., down-line loading). 6. Incremental updates; that is, it must explicitly indicate which components of the network may change without others also changing. The physical and administrative environmental controls shall be specified. Any assumptions about security of a given network should be clearly stated (e.g., the fact that all communications links must be physically protected to a certain level). The components of the network that form the NTCB must be identified. Furthermore, the modules within an NTCB par- tition that contain the reference validation mechanism (if any) within that partition must be identified. The procedures for the secure generation of a new ver- sion (or copy) of each NTCB partition from source must be described. The procedures and requirements for the secure generation of the NTCB necessitated by changes in the net- work configuration shall be described. Procedures for starting each NTCB partition in a secure state shall be specified. Procedures must also be included to resume secure operation of each NTCB partition and/or the NTCB after any lapse in system or subsystem operation. + Rationale There may be multiple system administrators with diverse responsibilities. The technical security measures described by these criteria must be used in conjunction with other forms of security in order to achieve security of the network. Additional forms include administrative security, physical security, emanations security, etc. Extension of this criterion to cover configuration aspects of the network is needed because, for example, proper interconnection of components is typically essential to achieve a correct realization of the network architec- ture. As mentioned in the section on Label Integrity, cryp- tography is one common mechanism employed to protect commun- ication circuits. Encryption transforms the representation of information so that it is unintelligible to unauthorized subjects. Reflecting this transformation, the sensitivity of the ciphertext is generally lower than the cleartext. If encryption methodologies are employed, they shall be approved by the National Security Agency (NSA). The encryption algorithm and its implementation are outside the scope of these interpretations. This algorithm and implementation may be implemented in a separate device or may be a function of a subject in a component not dedi- cated to encryption. Without prejudice, either implementa- tion packaging is referred to as an encryption mechanism herein. The requirements for descriptions of NTCB generation and identification of modules and components that form the NTCB are straightforward extensions of the TCSEC require- ments into the network context. In those cases where the vendor does not provide source code, an acceptable procedure shall be to request the vendor to perform the secure genera- tion. Given the nature of network systems (e.g., various com- ponents tend to be down at different times, and the network system must continue operation without that component), it is imperative to know both how to securely start up an NTCB partition, and how to resume operation securely. It is also necessary to know how to resume secure operation of the NTCB after any partition has been down. 4.1.4.3 Test Documentation + Statement from DoD 5200.28-STD The system developer shall provide to the evaluators a docu- ment that describes the test plan, test procedures that show how the security mechanisms were tested, and results of the security mechanisms' functional testing. It shall include results of testing the effectiveness of the methods used to reduce covert channel bandwidths. THE RESULTS OF THE MAP- PING BETWEEN THE FORMAL TOP-LEVEL SPECIFICATION AND THE TCB SOURCE CODE SHALL BE GIVEN. + Interpretation The "system developer" is interpreted as "the network system sponsor". The description of the test plan should establish the context in which the testing was or should be conducted. The description should identify any additional test components that are not part of the system being evaluated. This includes a description of the test-relevant functions of such test components and a description of the interfacing of those test components to the system being evaluated. The description of the test plan should also demonstrate that the tests adequately cover the network security policy. The tests should include the features described in the System Architecture and the System Integrity sections. The tests should also include network configuration and sizing. THE MAPPING BETWEEN THE FTLS AND THE NTCB SOURCE CODE MUST BE CHECKED TO ENSURE TO THE EXTENT POSSIBLE THAT THE FTLS IS A CORRECT REPRESENTATION OF THE SOURCE CODE, AND THAT THE FTLS HAS BEEN STRICTLY ADHERED TO DURING THE DESIGN AND DEVELOPMENT OF THE NETWORK SYSTEM. THIS CHECK MUST BE DONE FOR EACH COMPONENT OF THE NETWORK SYSTEM FOR WHICH AN FTLS EXISTS. + Rationale The entity being evaluated may be a networking subsys- tem (see Appendix A) to which other components must be added to make a complete network system. In that case, this interpretation is extended to include contextual definition because, at evaluation time, it is not possible to validate the test plans without the description of the context for testing the networking subsystem. The bandwidths of covert channels are used to determine the suitability of a network system for a given environment. The effectiveness of the methods used to reduce these bandwidths must therefore be accurately determined. 4.1.4.4 Design Documentation + Statement from DoD 5200.28-STD Documentation shall be available that provides a description of the manufacturer's philosophy of protection and an expla- nation of how this philosophy is translated into the TCB. The interfaces between the TCB modules shall be described. A formal description of the security policy model enforced by the TCB shall be available and an explanation provided to show that it is sufficient to enforce the security policy. The specific TCB protection mechanisms shall be identified and an explanation given to show that they satisfy the model. The descriptive top-level specification (DTLS) shall be shown to be an accurate description of the TCB interface. Documentation shall describe how the TCB implements the reference monitor concept and give an explanation why it is tamper resistant, cannot be bypassed, and is correctly implemented. The TCB implementation (i.e., in hardware, firmware, and software) shall be informally shown to be con- sistent with the FORMAL TOP-LEVEL SPECIFICATION (FTLS). The elements of the FTLS shall be shown, using informal tech- niques, to correspond to the elements of the TCB. Documen- tation shall describe how the TCB is structured to facili- tate testing and to enforce least privilege. This documen- tation shall also present the results of the covert channel analysis and the tradeoffs involved in restricting the chan- nels. All auditable events that may be used in the exploi- tation of known covert storage channels shall be identified. The bandwidths of known covert storage channels, the use of which is not detectable by the auditing mechanisms, shall be provided. (See the Covert Channel Guideline section.) HARDWARE, FIRMWARE, AND SOFTWARE MECHANISMS NOT DEALT WITH IN THE FTLS BUT STRICTLY INTERNAL TO THE TCB (E.G., MAPPING REGISTERS, DIRECT MEMORY ACCESS I/O) SHALL BE CLEARLY DESCRIBED. + Interpretation Explanation of how the sponsor's philosophy of protec- tion is translated into the NTCB shall include a description of how the NTCB is partitioned. The security policy also shall be stated. The description of the interfaces between the NTCB modules shall include the interface(s) between NTCB partitions and modules within the partitions if the modules exist. The sponsor shall describe the security architecture and design, including the allocation of security require- ments among components. The documentation includes both a system description and a set of component DTLS's. The system description addresses the network security architecture and design by specifying the types of components in the network, which ones are trusted, and in what way they must cooperate to support network security objectives. A component DTLS shall be provided for each trusted network component, i.e., each component containing an NTCB partition. Each component DTLS shall describe the interface to the NTCB partition of its component. Both the system description and each component DTLS shall be shown consistent with those assertions in the model that apply to it. Appendix A addresses component evaluation issues. To show the correspondence between the FTLS and the NTCB implementation, it suffices to show correspondence between each component FTLS and the NTCB partition in that component. As stated in the introduction to Division B, the spon- sor must demonstrate that the NTCB employs the reference monitor concept. The security policy model must be a model for a reference monitor. The security policy model for each partition implement- ing a reference monitor shall fully represent the access control policy supported by the partition, including the discretionary and mandatory security policy for secrecy and/or integrity. For the mandatory policy the single domi- nance relation for sensitivity labels, including secrecy and/or integrity components, shall be precisely defined. + Rationale The interpretation is a straightforward extension of the requirement into the context of a network system as defined for this network interpretation. Other documenta- tion, such as description of components and description of operating environment(s) in which the networking subsystem or network system is designed to function, is required else- where, e.g., in the Trusted Facility Manual. In order to be evaluated, a network must possess a coherent Network Security Architecture and Design. (Inter- connection of components that do not adhere to such a single coherent Network Security Architecture is addressed in the Interconnection of Accredited AIS, Appendix C.) The Network Security Architecture must address the security-relevant policies, objectives, and protocols. The Network Security Design specifies the interfaces and services that must be incorporated into the network so that it can be evaluated as a trusted entity. There may be multiple designs that con- form to the same architecture but are more or less incompa- tible and non-interoperable (except through the Interconnec- tion Rules). Security related mechanisms requiring coopera- tion among components are specified in the design in terms of their visible interfaces; mechanisms having no visible interfaces are not specified in this document but are left as implementation decisions. The Network Security Architecture and Design must be available from the network sponsor before evaluation of the network, or any component, can be undertaken. The Network Security Architecture and Design must be sufficiently com- plete, unambiguous, and free from obvious flaws to permit the construction or assembly of a trusted network based on the structure it specifies. When a component is being designed or presented for evaluation, or when a network assembled from components is assembled or presented for evaluation, there must be a priori evidence that the Network security Architecture and Design are satisfied. That is, the components can be assem- bled into a network that conforms in every way with the Net- work Security Architecture and Design to produce a physical realization that is trusted to the extent that its evalua- tion indicates. In order for a trusted network to be constructed from components that can be built independently, the Network Security Architecture and Design must completely and unambiguously define the security functionality of com- ponents as well as the interfaces between or among com- ponents. The Network Security Architecture and Design must be evaluated to determine that a network constructed to its specifications will in fact be trusted, that is, it will be evaluatable under these interpretations. The term "model" is used in several different ways in a network context, e.g., a "protocol reference model," a "for- mal network model," etc. Only the "security policy model" is addressed by this requirement and is specifically intended to model the interface, viz., "security perimeter," of the reference monitor and must meet all the requirements defined in the TCSEC. It must be shown that all parts of the TCB are a valid interpretation of the security policy model, i.e., that there is no change to the secure state except as represented by the model. Part II: Other Security Services ____ __ _____ ________ ________ 5. Introduction _ ____________ Part I of this Interpretation contains interpretations of the Department of Defense Trusted Computer System Evalua- tion Criteria (TCSEC), DOD 5200.28-STD. Part I deals with controlling access to information. Part II contains addi- tional network security concerns. These concerns differen- tiate the network environment from the stand-alone computer. Some concerns take on increased significance in the network environment; other concerns do not exist on stand-alone com- puters. Some of these concerns are outside the scope of Part I; others lack the theoretical basis and formal analysis underlying Part I. The criteria in this Part II address these concerns in the form of additional security requirements that may vary among applications. Overlap between Part I and Part II is minimized as much as possible. However, when an overlap occurs the association between the concerns addressed in both parts is defined. Part II ser- vices may be provided by mechanisms outside the NTCB. 5.1. Purpose and Scope _ _ _______ ___ _____ This Part II addresses network security disjoint from Part I. The rating derived in Part I is not effected by Part II. Every component or system must have a Part I evaluation as a basis for the Part II evaluation. Part II includes generic requirements, security features, and evaluation cri- teria. As described below, Part II evaluations differ from Part I. The purpose of these evaluations is similar, how- ever: to provide guidance to network managers and accredi- tors as to the reliance they can place in security services. These evaluations are input to the accreditor's decisions concerning the operational mode and range of sensitive information entrusted to the network. The network sponsor shall identify the security ser- vices offered by his system or component(s). Those services will be evaluated against the criteria for those services in Part II. 5.2. Criteria Form _ _ ________ ____ The general form of Part II criteria is a relatively brief statement, followed by a discussion of functionality, strength of mechanism, and assurance, as appropriate. Functionality refers to the objective and approach of a _____________ security service; it includes features, mechanism, and per- formance. Alternative approaches to achieving the desired functionality may be more suitable in different applications environments. Strength of mechanism refers to how well a specific ________ __ _________ approach may be expected to achieve its objectives. In some cases selection of parameters, such as number of bits used in a checksum or the number of permutations used in an encryption algorithm, can significantly affect strength of mechanism. Assurance refers to a basis for believing that the _________ functionality will be achieved; it includes tamper resis- tance, verifiability, and resistance against circumvention or bypass. Assurance is generally based on analysis involv- ing theory, testing, software engineering, validation and verification, and related approaches. The analysis may be formal or informal, theoretical or applied. For example, consider communications integrity protec- tion against message stream modification. A functionality decision is to select error detection only, or detection and correction; also one may select whether it is sufficient to detect an odd number of bit errors, error bursts of speci- fied duration, or a specified probability of an undetected error. Available mechanisms include parity, longitudinal redundancy check (LRC), cyclic redundancy check (CRC), and cryptographic checkfunction. The strength of the CRC is measured in the probability of an undetected error; this is dependent upon the number of bits employed in the CRC. There is no assurance of security associated with any of the mentioned mechanisms except cryptographic checkfunction. The algorithms are well known; an adversary could change message contents and recalculate the non-cryptographic checkfunction. The recipient would calculate the checkfunc- tion and not discover that the message had been manipulated. A cryptographic checkfunction would be resistant to such manipulation. 5.3. Evaluation Ratings _ _ __________ _______ Part II evaluations are qualitative, as compared with the hierarchically-ordered ratings (e.g., C1, C2, ...) resulting from Part I. At present it is not considered pos- sible or desirable to employ the same ratings scale in Part II. The results of a Part II evaluation for offered services will generally be summarized using the terms none, minimum, ____ _______ fair, and good. Services not offered by the sponsor will be ____ ____ assigned a rating of not offered. For some services it will ___ _______ be most meaningful to assign a rating of none or present. _______ The term none is used when the security service is not offered. In some cases the functionality evaluations may be limited to present or none. The assurance rating for each service is bounded by the Part I or Appendix A evaluation as appropriate because the integrity of the service depends on the protection of the NTCB. Table II-1 relates the Part II assurance rating to the minimum corresponding Part I evaluation ratings. These Part II evaluations tend to be more qualitative and subjective, and will exhibit greater variance than the Part I evaluations. Nevertheless, Part II evaluations are valuable information concerning the capabilities of the evaluated systems and their suitability for specific appli- cations environments. If functionality, strength of mechan- ism, and assurance are separately evaluated then a term may be applied to each. In some cases the strength of mechanism may be expressed quantitatively as a natural consequence of the technology (e.g., the number of bits in a CRC, the par- ticular function employed); this quantitative measure of strength may be employed as the basis for rating. The Part II evaluations may also be expected to exhibit a greater sensitivity to technological advances than the Part I evaluations. This sensitivity derives from the present empirical basis of some of the Part II security ser- vices as compared to the theoretical foundation of Part I. Research advances may help change this situation. As the state-of-the-art advances, the threshold for high evalua- tions may also be expected to increase. Therefore, a rating may become dated and may change upon reevaluation. In general, mechanisms that only protect against accidents and malfunctions cannot achieve an evaluation of strength of mechanism above minimum. Mechanisms must pro- vide protection against deliberate attacks in order to obtain at least a good evaluation. The summary report of a network product will contain the rating reflecting the Part I evaluation plus a paired list of Part II services and the evaluation for each. For example, network product XYZ might be rated as follows: [B2, security service-1: minimum, security service-2: not offered, security service-3: none, ... ,security service-n: (functionality: good, strength of mechanism: fair, assurance: good)]. In some cases where the security service is addressed outside this document (e.g., COMSEC), the evaluation from the external source may be reflected in the evaluation report. In such cases, the terms used will differ from those listed above. 5.4. Relationship to ISO OSI-Architecture _ _ ____________ __ ___ ___ ____________ An effort is underway to extend the ISO Open System Interconnection (OSI) architecture by defining "general security-related architectural elements which can be applied appropriately in the circumstances for which protection of communications between open systems is required." - Fami- liarity with OSI terminology is assumed in this discussion. The scope of this security addendum "provides a general description of security services and related mechanisms, _________________________ - ISO 7498/Part 2 - Security Architecture, ISO / TC ___ ____ ____ _ ________ ____________ 97 / SC 21 / N1528 / WG 1 Ad hoc group on Security, Project 97.21.18, September 1986. which may be provided by the Reference Model; and defines the positions within the Reference Model where the services and mechanisms may be provided." There is considerable overlap between the OSI Security Addendum and Part II. At the time of writing, the OSI docu- ment is evolving, making it difficult to exactly define the relationship. Therefore, the following statements may have to be modified in the future. Some of the security services identified in the OSI Security Addendum are covered by Part I of this Interpreta- tion; others are addressed in Part II. The emphasis is on making sure that all services are covered. The distinction between the security service and the mechanism that imple- ments the service is less strong in this Interpretation than in the OSI Security Addendum. The OSI Addendum generally addresses Functionality, occasionally addresses Strength of Mechanism, and rarely addresses Assurance, while in this Interpretation, especially in Part I, assurance is a major factor. The scope of the OSI Security Addendum is limited: "OSI Security is not concerned with security measures needed in end systems, installations and organizations except where these have implications on the choice and position of secu- rity services visible in OSI." The TCSEC and this Interpre- tation include OSI concerns as a proper subset. 5.5. Selecting Security Services for a Specific Environment _ _ _________ ________ ________ ___ _ ________ ___________ The enumeration of security services in Part II is representative of those services that an organization may choose to employ in a specific network for a specific environment. But not all security services will be equally important in a specific environment, nor will their relative importance be the same among different environments. The network management has to decide whether the rating achieved by a network product for a specific criterion is satisfac- tory for the application environment. As an abstract example, consider the network product XYZ which has received the rating [B2, security service-1: minimum, security service-2: not offered, ... ]. The management of network K may decide that they do not require security service-2, so the absence of this service does not effect the acceptability of the XYZ product; however, the management of network Q may decide that security service-2 is essential, so the absence of this service disqualifies product XYZ. The management of network P may decide secu- rity service-1 is very important and that any rating less than good is unacceptable, thereby disqualifying product XYZ; while the management of network R may decide that secu- rity service-1 need only be rated minimum. As a more concrete example, consider an application environment where wire-tapping is not a threat, such as aboard an airplane or in an underground bunker. A Local Area Network (LAN) in such an environment can be physically protected to the system-high security mode without encryp- tion because the system exists within a protected perimeter. In such environments, management may decide that labeling and access control based on labels provide sufficient pro- tection if sufficient mechanisms exist to protect the integrity of the labels. Cryptographic mechanisms are deemed unnecessary. By way of contrast, when the LAN environment involves passage through unprotected space, management may decide that a LAN must provide integrity pro- tection employing a cryptographic mechanism. 6. General Assurance Approaches _ _______ _________ __________ This section addresses assurance approaches applicable to many security services. The logic of the protocols and the implementation of countermeasures may be shown correct and effective by formal methods where possible (i.e., where tools exist) and infor- mal ones otherwise. To provide assurance that the security service can respond to various forms of external attacks, various methods of real and simulated testing can be applied, including: 1. Functional testing 2. Periodic testing 3. Penetration testing 4. Stress testing 5. Protocol testing for deadlock, liveness, and other security properties of the protocol suites In addition, the trusted computer base provides an exe- cution environment that is extremely valuable in enhancing the assurance of a variety of security services. The dis- cretionary and mandatory access controls can be employed in the design and implementation of these services to segregate unrelated services. Thus, service implementation that is complex and error-prone or obtained from an unevaluated sup- plier can be prevented from degrading the assurance of other services implemented in the same component. Furthermore, a TCB ensures that the basic protection of the security and integrity- of the information entrusted to the network is not diluted by various supporting security services identi- fied in this Part II. See also the discussion of Integrity in the Supportive Primitives section. In general, assurance may be provided by implementing these features in a limited set of subjects in each applica- ble NTCB partition whose code and data have a unique manda- tory integrity level to protect against circumvention and tampering. _________________________ - See, for example, Biba, K.J., "Integrity Considera- tion for Secure Computer Systems," ESD-TR-76-372, MTR- 3153, The MITRE Corporation, Bedford, MA, April 1977. Assurance of trustworthiness of the design and imple- mentation of Part II mechanisms may be related to the assurance requirements in Part I. The following factors are identified as contributing to an assurance evaluation: ser- vice design and implementation, service testing, design specification and verification, configuration management, and distribution. 6.1. Service Design and Implementation Factors _ _ _______ ______ ___ ______________ _______ An evaluation rating of fair indicates that the imple- mentation of the service employs the provisions of the TCB for a distinct address space. In addition, the implementa- tion of the service is internally structured into well- defined largely independent modules; makes effective use of available hardware to separate those elements that are protection-critical to the service from those that are not; is designed such that the principle of least privilege is enfor