Trust Management for Composite Services in Distributed Multi-agent Systems with Indirect Ratings: (Extended Abstract)

So far, trust evaluation models depend highly on collected evidences, which are mainly in the form of ratings, to produce trust values [3]. However, many situations require agents to form groups to provide more complicated services (composite services); meanwhile, consumers rate the quality of a composite service as a whole rather than individual group members [2]. Namely, all providers, who are in charge of the sub-services, are not rated individually. The lack of effective rating distribution mechanism has produced a gap in trust management systems since it causes the missing evidence of the members joined in composite services; thus, it can affect the robustness of trust evaluation models. To distribute a group rating is a non-trivial problem, especially when the given ratings could damage agents’ reputation. The subjective and indivisible nature of ratings restricts the use of fair-division [1] approaches in finding a solution. Under uncertainty assumption of members’ contribution and group structures, self-interest members could argue that they performed well and lay the blame on others for any undesirable rating of the group. To distribute group rating while reducing the conflict of interest of members, this paper proposes an evidence-based approach, considers supporting facts related to the performance of members to distribute the group rating. Firstly, we develop a communication protocol to cope with collecting evidence in distributed environment. We form the performance graph from the social relations of members and then distribute group ratings based on obtained graph characteristics.