A feedback double filtering based model for evaluating reputation in Peer-to-Peer networks

The feedback filtering is a key issue to the reputation evaluating in Peer-to-Peer networks. The existing models usually focus on filtering out the fake feedback at the trustor´s side. However, even the most honest recommender could submit feedbacks with low quality. In this paper, a new reputation evaluating model is proposed. The model filters the feedback not only at the recommender´s side but also at the trustor´s side. Two new measures, i.e. confidence degree of feedback and trustworthiness degree of recommending, are introduced to the model. The former represents the confidence of the recommender to its feedback. This measure can be used by the recommender to filter out the feedback with low confidence. The latter is used to weight the recommender´s feedback at the trustor´s side. Experimental results show that our model has better performance and is robust even with large amount of malicious peers.