基于连续行为观察的P2P网络中邻居评价模型

基于荣誉的信任机制是对P2P网络节点行为进行评价的重要手段，用来保证P2P网络应用的健康进行.信任机制在对一个节点进行评价时需要获得其他节点的局部信任值信息.目前局部信任值的计算由于不考虑策略节点和人类评价误差两种重要因素的影响，难以准确反映网络节点的特征.提出了一种P2P网络中邻居行为的评价模型PeerStrategy,该模型使用确定的有限状态机(DFA)对邻居连续行为的状态变化进行刻画.通过关注邻居在任意连续行为中引起负面评价的概率，既能够较为准确地发现网络中的策略节点，又能够容忍一定程度的人类评价误差.仿真实验表明，该模型显著提高了局部信任值的准确度，并降低了对全局信任值估计误差影响，明显优于当前的其他局部信任值计算方法.

A Consecutive-Behaviors-Observing-Based Neighbor Evaluation Model in P2P Network

Reputation based trust mechanism has been identified as an effective method to evaluate peers behavior and which is employed to secure the applications in P2P network. Trust mechanism is such a mechanism that relies on other peers reports, which are also called local trust, to evaluate a designated peer. However, the existence of strategic peers and human judgment error is a big challenge, which makes the local trust hard to reflect peers type. Furthermore, it increases the estimation error of global trust. The authors propose a new model, called PeerStrategy, to evaluate neighbors behavior in P2P network. This model explores deterministic finite automaton (DFA) to describe the variance of neighbors consecutive behaviors. The DFA consists of seven states and it transits between states by neighbors performance in the interactions. By examining the probability of negative behaviors in any consecutive ones, the model can not only detect strategic peers accurately but also tolerate human judgment error. As a result, this model improves the accuracy of local trust, and whats more, it decreases the estimation error of global trust. The simulation shows that this model improves the accuracy of local trust considerately and also diminishes the influence on the estimation error of global trust, and it performs the best compared with other current methods.