Access Control (AC) systems are among the most critical of network security components. A system’s privacy and security controls are more likely to be compromised due to the misconfiguration of access control policies rather than the failure of cryptographic primitives or protocols. This problem bec

Attribute relations in access control mechanisms or languages allow accurate and efficient specification of some popular access control models. However, most of the access control systems including today's de-facto access control protocol and specification language, XACML, does not provide sufficient syntactic and semantic support for the specification of attribute relations in their scheme. In this paper, we show the deficiencies of XACML in specifying such capabilities in the implementations of the Multilevel Security, Hierarchical Role Based policies and Separation of Duty requirements of access control systems. In comparison, we then demonstrate the attribute relation mechanism provided by a relation-based access control mechanism - the Policy Machine.

Security policy enforcement is instrumental in preventing the unauthorized disclosure of sensitive data, protecting the integrity of vital data, mitigating the likelihood of fraud, and ultimately enabling the secure sharing of information. In accessing a given resource, policy may dictate, for example that a user has a need-to-know, is appropriately cleared, is competent, has not already performed a different operation on the same resource, the resource was previously accessed by a different user, is incapable of accessing other enterprise resources, or is capable of accessing an object or any copy of the object while performing a specific task. Currently, there exist a rich set of formal security models that can translate organizational policies. A small sample of well documented policies include, avors of Discretionary Access Control (DAC), Mandatory Access Control (MAC), Role-Based Access Control (RBAC), ORCON, Chinese wall, and History-Based Separation of Duty. Enterprise policies that are designed to protect resources are also ad-hoc in nature.

Many researchers have tackled the architecture and requirements aspects of grid security, concentrating on the authentication or authorization mediation instead of authorization techniques, especially the topic of policy combination. Policy combination is an essential requirement of grid, not only because of the required remote (or global) vs. local interaction between grid members, but also the dynamic scalability nature of handling the joining and leaving of grid membership. However, evolving from the general security requirements of grid, the independency of a grid member's access control system is critical and needs to be maintained when the access decision is determined by the combination of global and local access control policies. The Policy Machine (PM) provides features which not only can meet the significant independency requirement but also have better performance, easier management, and more straightforward policy expression than most of the popular policy combination techniques for grid.

As a major component of any host, or network operating system, access control mechanisms come in a wide variety of forms, each with their individual attributes, functions, methods for configuring policy, and a tight coupling to a class of policies. To afford generalized protection, NIST has initiated a project in pursuit of a standardized access control mechanism, referred to as the Policy Machine (PM) that requires changes only in its configuration in the enforcement of arbitrary and organization specific attribute-based access control policies. Included among the PM's enforceable policies are combinations of policy instances (e.g., Role-Based Access Control and Multi-Level Security). In our effort to devise a generic access control mechanism, we construct the PM in terms of what we believe to be abstractions, properties and functions that are fundamental to policy configuration and enforcement. In its protection of objects under one or more policy instances, the PM categorizes users and objects and their attributes into policy classes, and transparently enforces these policies through a series of fixed PM functions, that are invoked in response to user or subject (process) access requests.

Role-based Access Control (RBAC) models have been implemented not only in self-contained resource management products such as DBMSs and Operating Systems but also in a class of products called Enterprise Security Management Systems (ESMS). ESMS products are used for centralized management of authorizations for resources resident in several heterogeneous systems (called target systems) distributed throughout the enterprise. The RBAC model used in an ESMS is called the Enterprise RBAC model (ERBAC). An ERBAC model can be used to specify not only sophisticated access requirements centrally for resources resident in several target systems, but also administrative data required to map those defined access requirements to the access control structures native to the target platforms. However, the ERBAC model (i.e., the RBAC implementation) supported in many commercial ESMS products has not taken full advantage of policy specification capabilities of RBAC. In this paper we describe an implementation of ESMS called the 'Role Control Center' (RCC) that supports an ERBAC model that includes features such as general role hierarchy, static separation of duty constraints, and an advanced permission review facility (as defined in NIST's proposed RBAC standard). We outline the various modules in the RCC architecture and describe how they collectively provide support for authorization administration tasks at the enterprise and target-system levels.

The question before the panel: Considering all factors (for example: quality of protection, performance, compatibility, ease of use), which operating system access control technique will provide the greatest overall benefit to users?

In this article we propose a standard for role-based access control (RBAC). Although RBAC models have received broad support as a generalized approach to access control, and are well recognized for their many advantages in performing large-scale authorization management, no single authoritative definition of RBAC exists today. This lack of a widely accepted model results in uncertainty and confusion about RBAC's utility and meaning. The standard proposed here seeks to resolve this situation by unifying ideas from a base of frequently referenced RBAC models, commercial products, and research prototypes. It is intended to serve as a foundation for product development, evaluation, and procurement specification. Although RBAC continues to evolve as users, researchers, and vendors gain experience with its application, we feel the features and components proposed in this standard represent a fundamental and stable set of mechanisms that may be enhanced by developers in further meeting the needs of their customers. As such, this document does not attempt to standardize RBAC features beyond those that have achieved acceptance in the commercial marketplace and research community, but instead focuses on defining a fundamental and stable set of RBAC components. This standard is organized into the RBAC Reference Model and the RBAC System and Administrative Functional Specification. The reference model defines the scope of features that comprise the standard and provides a consistent vocabulary in support of the specification. The RBAC System and Administrative Functional Specification defines functional requirements for administrative operations and queries for the creation, maintenance, and review of RBAC sets and relations, as well as for specifying system level functionality in support of session attribute management and an access control decision process.

Existing software infrastructures and middleware provide uniform security services across heterogeneous information networks. However few, if any, tools exist that support access control policy management for and between large enterprise information networks. Insiders often exploit gaps in policies to mount devastating attacks. This paper presents a Policy Machine and Policy Mediation Architecture for coordinating diverse policies in large information networks. The language-based approach adopted by each of these technologies permits local and global access control policy validation with static analysis and other formal techniques. Together the Policy Machine and Policy Mediation Architecture comprise an effective system for closing policy gaps in multi-enterprise environments.

This paper describes NIST's enhanced RBAC model and our approach to designing and implementing RBAC features for networked Web servers. The RBAC model formalized in this paper is based on the properties that were first described in Ferraiolo and Kuhn [1992] and Ferraiolo et al. [1995], with adjustments resulting from experience gained by prototype implementations, market analysis, and observations made by Jansen [1988] and Hoffman [1996]. The implementation of RBAC for the Web (RBAC/Web) provides an alternative to the conventional means of administering and enforcing authorization policy on a server-by-server basis. RBAC/Web provides administrators with a means of managing authorization data at the enterprise level, in a manner consistent with the current set of laws, regulations, and practices.

Formally defines a wide variety of separation-of-duty (SoD) properties, including the best known to date, and establishes their relationships within a formal model of role-based access control (RBAC). The formalism helps to remove all the ambiguities of informal definition and offers a wide choice of implementation strategies. We also explore the composability of SoD properties and policies under a simple criterion. We conclude that the practical implementation of SoD policies requires new methods and tools for security administration, even within applications that already support RBAC, such as most database management systems.