A security infrastructure for trust management in multi-agent systems

Multi-agent systems are based on the interaction of autonomous software components, the agents, which cooperate to achieve common goals. But such interactions, just as in human societies, can be set correctly only on the base of trust relations. This paper presents a security model founded on delegation certificates, which allows the management of security policies on the base of trust relations among autonomous software agents, collaborating and competing in wide, open and evolving agent societies. On the converse, the presence of a strong and flexible security infrastructure is fundamental to develop trust in counterparts and start advantageous interactions with them. While some of these concepts are already being adopted into the development of the security layer for JADE, a standard-based and widely deployed framework to build multi-agent systems, this paper includes further ideas which distributed multi-agent security frameworks could benefit from.