A Context-aware Computational Trust Model for Multi-agent Systems

Context information is taken into account when evaluating agents´ trust to reflect the dynamic feature of multi-agent systems in open uncertain environment more accurately. A context-aware computational trust model based on Bayesian networks is proposed. Unlike many existing intuitive solutions, both the reputation estimation and unfair recommender filtering use statistical approaches. Risk is explicitly combined with trust to help agents making proper decisions. Simulation results show that our prototype system can make accurate trust inferences and is robust against unfair recommenders.