A database authorisation framework for multi-agent systems

In a database environment, the authorisation paradigm is recognised as an effective way to achieve security and privacy of information read/written by multiple users. In this paper we present a database authorisation framework for multi-agent systems, and show that to manage access to the database through which agents coordinate their actions, authorisations must reflect the semantics of agents´ cooperation. The framework incorporates non-recursive grant and revoke operations so that when a privilege is added/removed from an agent accessing a shared cooperation entity all authorisations granted by that agent are inserted/deleted in an efficient manner. We show how agents administrate their joint relations, deduction rules and objects using a cooperation domain (i.e. an event-driven distributed multibase) to support consistency and fault-tolerance. We also address the management of views handled by agents in support of such cooperation domains