Decentralized Intrusion Detection for Secure Cooperative Multi - Agent Systems

In this paper, the authors address the problem of detecting faulty behaviors of cooperative mobile agents. A novel decentralized and scalable architecture that can be adopted to realize a monitor of the agents' behavior is proposed. They consider agents which may perform different independent tasks, but cooperate to guarantee the entire system's safety. Agents plan their next actions by following a set of rules which is shared among them. Such rules are decentralized, i.e., they depend only on configurations of neighboring agents. Some agents may not be acting according to this cooperation protocol, due to spontaneous failure or tampering. To detect such misbehaviors they propose a solution where each agent runs a local observer using only locally available information.