The Why and How of Delegations in Distributed Security Systems

In this paper we address the main questions associated with delegations and the role they play in security architectures for distributed systems; why they are required and how they can be implemented. In distributed and dynamic systems like the Semantic Web or pervasive computing environments, describing comprehensive policies in terms of every possible resource and user is not always possible since neither resources nor users can be completed enumerated a-priori. Even if all users and resources can be pre-determined, it is not feasible to develop complete policies because it is not possible to account for every access right or task required in these dynamic environments. We believe that policies should be as simple as possible and control should be decentralized, that is authorization should be possible by more than just a few key entities and tasks/responsibilities should be delegate-able. We believe that delegations are the key as they allow policies to be less exhaustive and allow policies to be dynamically updated. We address two kinds of delegations: delegation of rights and delegation of responsibility. Delegation of rights allows access rights to be transferred whereas delegation of responsibility allows obligations to be transferred. We describe our policy language, which is based on deontic concepts and speech acts, that allows both these kinds of delegations to be represented and reasoned over. Speech acts like delegation, revocation, request and cancel are an integral part of the design as they provide decentralized control by allowing policies to be dynamically modified. We discuss the delegation protocols that we have developed and describe our integrated approach to policies and speech acts. We describe our prototype implementation of a security system, which supports decentralized control through policies and delegation statements represented in logic and a semantic language, Resource Description Language Schema.