Abstract: To solve the problems of dynamic characteristic, and deal with collusion attack in current P2P trust model, a new dynamic comprehensive trust model（DOGC trust） is proposed. It shares node equally to different group competition in logical based on global trust in this model. Weight and entropy are counted, local trust of node is updated, local trust’s standard deviation in group and node is calculated based on entropy method. Node is adjusted by dynamic optimization. That not only leads group’s weight in to the model for increasing the calculation accuracy, but also pays attention the similarity of node's behavior to strengthen the incentives strategy. The simulation and experimental results indicate that the new model has advantages in controlling dynamic characteristic of node, collusion attack and improving transaction success rate over the existing trust models.

Key words: Peer-to-Peer（P2P） network, dynamic characteristic, collusion attack, entropy method, trust

摘要： 针对P2P网络中节点行为的动态性改变和共谋攻击等关键问题，提出了一种基于熵权的群组动态优化竞争的P2P信任模型。该模型以节点全局信任度为依据，在逻辑上将节点分至不同速率群组；基于“熵权法”计算群组熵值及权重值，并更新时间周期内群组节点局部信任度，计算群组局部评价信誉标准差和节点局部评价标准差；采用动态优化竞争策略，调整节点或群组。该模型不仅在节点分组的基础上引入了群组权重值，以提升信任计算的准确度，而且在优化竞争过程中考虑节点行为的相似性，以强化激励策略。仿真实验表明，该模型在识别、抑制节点的动态性振荡行为和抑制共谋攻击行为方面均有一定的改进，在交易成功率方面有一定的提高。

关键词: 点对点（P2P网络）, 动态性, 共谋攻击, 熵权法, 信任