Abstract: This article proposes a novel trust models based on the recent behavior of peers to solve the problem of the large computation and dynamics in the existing trust model.The recent fail ratio of one peer records its recent state.The reputation of one malicious peer will fall to 0，when the node has vicious behavior，then the course of accumulated the trust will become slow.The response of the system changes fast，because it adopts the posterior method and cuts down the computations before the transaction.Theoretical analysis and simulation show that the models use the trust of dynamic computation，and meet the order of small computation，prevent the mutual behavior efficiently.

Key words: peer-to-peer, trust model, security, recent behavior

摘要： 针对现有动态信任模型计算量大的问题，给出了基于近期表现的P2P网络信任模型。利用近期失败率来记录与该节点交易的最近状态，当某节点突然“爆发”恶意本性时，可以很快地将其信任值降到0，当某节点有过恶意行为后，再去积累信任值，积累过程会变慢。由于采用信任值的后验方式，减少了交易前的计算量，提升了系统的反应速度。仿真实验分析表明，该模型提供了信任值计算的动态性，同时满足了小计算量的要求，而且可以有效防止共谋行为。

关键词: P2P网络, 信任模型, 安全, 近期表现