基于Dirichlet分布的无线传感器网络的信誉计算模型研究

Reputation Computing for Wireless Sensor Networks Based on Dirichlet Distribution

提出了基于无线传感器网络的信任评估框架,在该框架中,分为检测模块和信誉模块两部分,其中检测模块用于检测各节点测量的数据的异常程度,并根据此异常程度将节点行为分为三大类;在信誉模块中,建立基于Dirichlet分布的信誉函数,计算节点的可信度及不确定度,并实时地利用新获得的数据对信誉进行更新;根据其他节点的信誉进行整合信誉,加快了收敛速度。仿真的结果表明,该模型可以检测出多种错误节点,并且与传统的基于β-分布的信誉模型相比,更加适合于无线传感器网络。

This paper presents a reputation computing framework that based on wireless sensor network. In the framework , there are two modules （detection and reputation module）. The detection module determines each node＇s outlier degree using the LOF outlier detection algorithms, and according with the outlier degree, the nodes will be divided into three classes. The reputation module includes obtaining the reputation information of its neighbors, establishing the reputation function based on Dirichlet distribution, and accounting the belief value and uncertainty value. Further more, the system can update the node＇s reputation real-time using the new data, and integrate the reputation according on other nodes＇ reputation information, so the convergence speed can be increased. The result of simulation indicates that the framework can detect manifold faulty nodes, and being compared with the reputation framework that based on beta distribution, this framework is more suitable for wireless sensor network.