基于K-means算法的无线传感器网络节点自私行为检测方法  

Detection Method for Nodes Selfish Behavior of Wireless Sensor Networks Based on K-means Algorithm

作　　者：陈波[1] 毛剑琳[1] 郭宁[1] 乔冠华[1] 戴宁[1] 

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650000

出　　处：《系统仿真学报》2014年第3期580-585,共6页Journal of System Simulation

基　　金：国家自然科学基金资助项目(61163051);云南省应用基础研究基金资助项目(2009ZC050M)

摘　　要：针对无线传感网络共享信道中自私节点(或恶意节点)对信道的不公平竞争行为,提出了一种基于网络性能特征序列的聚类检测方法(Network Performance Characteristic Sequence based-Clustering Detection Method,NPCS-CDM)。该算法以节点链路的平均传输延迟和平均吞吐量为网络性能特征建立统计序列,采用K-means聚类算法对特征序列进行分析和聚类,以此完成网络中节点自私行为的检测,同时该方法有效解决了基于CUSUM算法用于检测多自私节点的不足,即难以确定适当的阈值来完成检测任务。基于NS2的仿真结果表明,NPCS-CDM对自私节点的检测效果明显优于已有的基于CUSUM的算法,而且能适用于多自私节点存在的情况。Since there is unfair competition resulted by selfish nodes/ hostile nodes within a shared channel in wireless sensor networks, a Network Performance Characteristic Sequence based-Clustering Detection Method （NPCS-CDM） for detection of selfish behavior was proposed. The average delay and the average throughput of each node link were taken as the network performanee characteristic, and then statistic sequence was established, based on which K-means clustering algorithm was used to classify the data and distinguish the selfish behavior from the normal behavior. The method is superior to CUSUM algorithm, where it is difficult to determine an appropriate threshold to distinguish multiple selfish nodes from the network. Simulation experiments based on NS2 indicate that the NPCS-CDM algorithm can effectively detect both one selfish node and the multiple selfish nodes from the network.

关 键 词：无线传感器网络 自私节点 特征序列 K-MEANS 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]