基于组合加权k近邻分类的无线传感网络节点复制攻击检测方法

Detection of Node Replication Attacks in Wireless Sensor Networks Based on Combined Weighted k-Nearest Neighbor Classification

无线传感网络节点体积小,隐蔽性强,节点复制攻击检测的难度较大,为此提出一种基于组合加权k近邻分类的无线传感网络节点复制攻击检测方法。通过信标节点的空间位置数据与相距跳数得出各节点之间的相似程度,结合高斯径向基核函数求解未知节点的横轴、纵轴的空间坐标,确定各网络节点的空间位置;根据网络节点的属性特征与投票机制建立节点复制攻击模型,凭借组合加权k近邻分类法划分节点类型,并将结果传送至簇头节点,由簇头节点做出最后的仲裁,识别出节点复制攻击行为。仿真结果表明,所提方法的节点复制攻击检测率最大值为99.5%,最小值为97.9%,对节点复制攻击检测的耗时为5.41 s,通信开销数据包数量最大值为209个,最小值为81个。

Wireless sensor network nodes have small size and strong concealment,so it is difficult to detect node replication attacks.The similarity between nodes is obtained according to the spatial position data of beacon nodes and the number of hops apart.The spatial co-ordinates of the horizontal axis and vertical axis of unknown nodes are solved by combining the Gaussian radial basis kernel function to determine the spatial position of each network node.The node replication attack model is established according to the attribute charac-teristics and voting mechanism of network nodes.The node types are divided by using the combined weighted k-nearest neighbor classifi-cation method,and the results are transmitted to the cluster head node,which makes the final arbitration to identify the node replication attack behavior.The simulation results show that the maximum detection rate of node replication attack is 99.5%and the minimum de-tection rate is 97.9%.The detection time of node replication attack is 5.41 s.The maximum number of communication overhead packets is 209 and the minimum number is 81.