基于最小计数概要的无线传感器网络异常检测方案

摘要

A count-min sketch based anomaly detection scheme is proposed,which is suited to hierarchical wireless sensor net-works.In the proposed scheme,we first partition the whole sensor network into several clusters in which the cluster members are physically adjacent and data-correlated.Then,in the cluster head,the count-min sketches are collected from each cluster member and compared with its own count-min sketch in the form of Kullback-Leibler divergence.We show through experiments that the proposed scheme has a higher detection accuracy ratio and a lower false alarm ratio than the existed anomaly detection schemes. Meanwhile,it requires less storage space and consumes less energy than the non-sketch based anomaly detection scheme.%针对层次结构的无线传感器网络，提出了一种基于最小计数概要的异常检测方案。首先，将整个无线传感器网络划分成多个簇，使得簇内每个传感器节点位置临近且具有相似的感知数据；其次，由簇头节点收集成员节点感知数据的最小计数概要，并以 Kullback-Leibler 距离的方式与自身的最小计数概要进行匹配，从而识别异常节点。实验结果表明，与现有异常检测方案相比，所提出的方案具有较高的检测精度和较低的误报率，并且所需存储空间和传输能量消耗较低。