基于分簇的传感器网络异常检测算法

Outlier detection algorithm based on clustering in sensor network

针对传感器网络中集中式异常检测算法能耗过高的问题,提出了基于分簇的传感器网络异常检测算法.对网络进行分簇,簇内节点将采样数据传至簇头,簇头节点保持节点的部分历史数据并进行节点间相似性比较,没有达到给定支持度的数据就是异常数据.对于异常数据可以直接丢弃,不用传回基站以节约能量开销.该算法在网内对传感器网络进行数据处理,避免将所有采样数据传往基站.实验结果表明:该算法可以有效节约节点能量开销,延长传感器网络生命周期.

Aiming at the over high energy consumption of centralized outlier detection algorithm in sensor network,a distributed outlier detection algorithm based on cluster was proposed.The sensor network was divided into some clusters.A cluster head was selected for each cluster.Common sensor nodes send sample data to their cluster heads.Cluster heads maintained some history data and executed correlation similarity comparison.Data which had not enough supporting will be regarded as outlier data.Outlier data will be discarded and need not send to base station to save energy.The algorithm avoided sending all sample data to base station by in-network data processing.Experiment results show that the algorithm can effectively save energy consumption and improve the lifetime of sensor network.