Abstract: According to the existing P2P network's deficiency of inadequately considering the difference and the dynamic in peers' behavior when aggregating peer trust valucs,a new dynamic comprehensive trust model (DWAI}rust) was proposed. In this model, a Multi-dimensional Time Series are firstly applied to record evaluations on peers in some past time segments responding to different indexes. Secondly, the different weight for each evaluation index in each time segment is computed by "Entropy Method" and then the first comprehensive weighted evaluation result in different time segment can he obtained. Finally, the time weight for different time segment can be computed by resolving a nonlinear programming problem and then the second comprehensive weighted evaluation is carried out to obtain the final trust value. Hence, this model not only takes into the consideration that the dynamic changes in transaction context and peer context will influent the peer trust value, but also employs time granularity to reflect the changes. Simulation experiment shows that the model can well identify the malicious peer behavior such as taking strategy to proceed periodic oscillation cheating and then improve the improve transaction success rate.

Key words: P2P,Dynamic comprehensive evaluation,Doubly weighted average,Multi-dimensional time series