Using Trustworthy and Referee Agents to Secure Multi-Agent Systems

Abstract

Security is a crucial factor in multi-agent systems where agents dynamically leave and enter the system. In this paper, we continue our work on security for agent-based systems by proposing a new trust model allowing agents to decide about a target agent. The model combines two techniques: using reports provided by the trustworthy agents regarding to direct and indirect interaction histories, and using reports provided by the referee agents in the form of recommendations. In addition, after a periodic time, the actual behavior of the target agent is checked against the provided information by others to adjust their credibility. The paper considers different parameters for computing trust, which is represented as a quantitative value.