基于中值滤波的分布式扩散最小均方算法

An improved distributed diffusion least mean square algorithm

研究无线传感网潜在的恶意攻击节点对系统整体估计性能的影响,从节点选择的角度入手,改进原有扩散最小均方算法的融合步骤。针对接收到的相邻居节点估计值采用中值滤波处理,替代原来的融合方法,以降低恶意攻击节点对整个网络的影响。通过改变恶意节点攻击强度或恶意节点度,对改进算法进行仿真,结果显示全局均方偏差值有所降低,且改进算法对恶意节点攻击强度和恶意节点度的改变并不敏感。

To study the impact of malicious nodes on signal estimation performance for wireless sensor networks （WSN） in an untrustworthy environment, the diffusion least mean square （MFDLMS） algorithm based on the median filtering mechanism is proposed from the perspective of the node selection. The existing fusion step of the diffusion least mean （DLMS） algorithm is therefore improved to reduce the impact of malicious attacks on the entire network, of which the estimated value received from neighbor nodes are processed by the median filtering. The improved algorithm is simulated with the change of the attack power and the degree of malicious nodes. Results show that the mean square deviation （MSD） curve is decreased, which means that the improved algorithm is insensitive to the attack power and the degree of malicious nodes.