Abstract: In this paper，it presents a distinguishability of penalty based trust model，named DPTrust，in which the duration of penalty is determined in terms of peers credility level in a latest time window，as a result，both kinds of peers has been distinguished．Mathematic analysis and simulation experiments show that，the proposed model can improve incentives for peers and total payoffs of P２P systems effectively．

Key words: Peer－to－Peer network, reputation model, repeated games, penalty strategy

摘要： 提出了一种具有惩罚水平区分度的重复博弈信任管理模型DPTrust，对节点的非合作行为，依据其在最近时间窗口中的信用级别，施加不同时长的惩罚期，从而区分了善意节点的“偶尔犯错”和恶意节点的“故伎重演”及摇摆等行为。理论分析和仿真实验表明，该系统有效地提高了P２P节点诚实交易的积极性和系统的整体效率。

关键词: 对等网络, 信誉模型, 重复博弈, 惩罚策略