Research of Trust Model based on Interest Group and Similarity Recom- mendation

Trust should be substantially based on evidences including historical behaviors and recommendations in p2p. But experiments show that nodes usually just focus on the services they are interested in. For others they may ignore or feedback at will. So in most cases the system suffers a lot from dishonest feedbacks and strategically cheating behaviors of malicious nodes. Paper proposes a new trust model based on interest group and similarity recommendation, in which nodes belong to their interesting groups, give feedback only to the service supplied by the same group for improving the accuracy of the feedback and recommendations. By this way nodes trust are divided into two parts: trust intra-group and trust inter-group according to their relations. Experiment results show that our trust model can help nodes effectively select appropriate interaction partners.