Abstract: In P2P, trust is the basic of effective interactions between nodes but because of the anonymity and dynamic characters, it is difficulty to establish and maintain the trust relationships in the network. In fact trust is not objective but has subjectivity and ambiguity, so this study gives a new distributed trust model based on fuzzy sets theory for evaluating the trust values of nodes which considers multiple network factors, different weights, different interaction results and finally proposes a general algorithm for calculating nodes trust. Specifically, this trust model incorporates both direct trust and indirect trust and provides a flexible method to combine the two different parts. At last, by simulations we can see that this trust model can distinguish good nodes and malicious nodes, and give great help for nodes to effectively select appropriate interaction partners.