基于信誉机制的分布式扩散最小均方算法

Distributed Diffusion Least Mean Square Algorithm Based on the Reputation Mechanism

非安全环境中的无线传感器网络(WSN)可能存在恶意攻击节点,恶意节点将会篡改其观测数据以影响参数估计的准确性。为此,该文提出基于信誉机制的分布式扩散最小均方(R-d LMS)算法和扩散归一化最小均方(R-d NLMS)算法。该算法能够根据各节点对整个网络参数估计的贡献来设置相应的信誉值,从而减小恶意节点对网络攻击的影响。仿真结果表明,与无信誉值的算法相比,该算法的性能得到大幅度提高,且R-d NLMS算法在R-d LMS算法的基础上,算法性能得到进一步提升。

To deal with the problem of signal estimation for Wireless Sensor Networks （WSN） in a untrustworthy environment where malicious nodes tamper the measured data, two reputation-based Mgorithms, that are, Reputation-based diffusion Least Mean Square （R-dLMS） algorithm and Reputation-based diffusion Normalized Least Mean Square （R-dNLMS） algorithm, are proposed. The proposed algorithms could assign the appropriate reputation value to each node according to its contribution to the whole network, and minimize the reputation value of malicious nodes to lower the impact of malicious nodes in the network. Simulation results show that the proposed algorithms can greatly improve the performance compared with the one without reputation value, and the performance of R-dNLMS algorithm has been further improved based on R-dLMS algorithm.