对等网文件共享系统高度的开放性和自治性导致节点在交互的过程中易受诸如欺骗,协同作弊等安全威胁.针对节点在交互过程可能出现的安全问题提出一个基于贴近度加权推荐的全局信任模型MGTrust,根据推荐者自身对其他节点评价贴近程度分配权值,以此评价节点的可信度,给出信任模型的分布式实现,最后对模型实验分析,结果表明MGTrust模型与传统模型相比在迭代速度及成功下载率等指标上有明显改善. The high degree of openness and autonomy of peer-to-peer file sharing system makes nodes susceptible to security threats such as cheating, cheating and cheating in the process of interaction.Aiming at the possible security problems during the interaction of nodes, a global based on close-degree weighted recommendation Trust model MGTrust, according to the recommender itself to other nodes evaluation of the degree of proximity to assign weights, in order to evaluate the credibility of the node to give the trust model of the distributed implementation, and finally model experiments, the results show that MGTrust model with the traditional model phase Than in the iteration speed and success rate and other indicators have significantly improved.