Model for Malicious Node Recognition Based on Environmental Parameter Optimization and Time Reputation Sequence

基于环境参数优化和时间信誉序列的恶意节点识别模型

¹ Northeast Electric Power University, Key Laboratory of Modern Power System Simulation and Control&Renewable Energy Technology, Ministry of Education, Jilin 132012, China
² School of Electrical Engineering, Northeast Electric Power University, Jilin 132012, China
³ School of Automation Engineering, Northeast Electric Power University, Jilin 132012, China
⁴ Ericsson(Xi'an) Information and Communication Technology Service Co., Ltd. Dalian Branch, Dalian 116000, China

Received: 8 July 2019

Abstract

Wireless sensor network (WSN) works in a complex environment. To interdict the malicious nodes which attacks the safety of network, such as interrupt attacks and selective forwarding attacks, based on TS-BRS reputation model, a model for malicious node identification based on MNRT-OEP&RS algorithm is constructed. Using the linear regression of machine learning and combining the energy of nodes, data volume, number of adjacent nodes, the node sparsity and other deterministic parameters can solve environmental parameters. Then the similarity of between the benchmark reputation sequence and cycle reputation sequence sets the dynamic reputation double threshold are calculated in order to identify the malicious nodes by dynamically considering the information forwarding behavior. The simulated results show that the improved algorithm can guarantee the security of wireless sensor networks in complex environments effectively with above 90% recognition of malicious nodes and below 8% false positive rate.

摘要

在复杂无线传感器网络环境中，为阻断恶意节点发动危及网络安全的中断攻击和选择性转发攻击，在TS-BRS信誉模型的基础上，搭建基于MNRT-OEP&RS的恶意节点识别模型，利用机器学习中的线性回归并结合节点能量、工作量、邻节点数量、节点疏松度等可确定参数求解环境参数，计算基准信誉序列与周期内的节点信誉序列的相似度；设定动态信誉双阈值，对节点的信息转发行为进行动态考量，以甄别恶意节点。仿真实验表明，改进后的算法对恶意节点的识别率可达90%以上，对正常节点误判率降低到8%以下，有效提高复杂环境下无线传感器网络的安全性。