基于机器学习的无线传感器网络入侵检测算法

Machine learning-based intrusion detection technology for wireless sensor networks

针对资源受限的无线传感器网络入侵检测效果不佳的问题,本文提出了一种基于机器学习的无线传感器网络入侵检测算法。该算法将数据局部密度和数据特征距离引入模糊聚类,提高了聚类有效性的同时降低了聚类收敛的时间;该算法将聚类得到的模糊隶属度作为模糊因子应用于模糊支持向量机,降低了人为选取模糊因子造成的主观性,将噪声点和孤立点对分类造成的影响降至最低。以WSN-DS数据集为实验数据,理论分析及实验结果表明:本文提出的入侵检测算法具有检测率高、计算复杂度低等特性,能够适用WSNs这一应用场景。

This study aims to optimize the resource-limited wireless sensor networks(WSNs)intrusion detection algorithm proposing an algorithm based on machine learning.First,the algorithm introduces the data of local density and feature distance into a fuzzy clustering method,which improves the clustering effectiveness and reduces the time of cluster convergence.Second,the algorithm applies the fuzzy membership degree,which is obtained by the improved fuzzy clustering method,as the fuzzy factor to the fuzzy support vector machine to reduce the subjectivity caused by the artificial selection of the fuzzy factor and minimize the impact of the noise and isolated points on the classification.The WSN-DS dataset is used as an experimental dataset.The theoretical and experimental results show that the proposed intrusion detection algorithm is characterized by a high detection rate and a low computational complexity,and can be applied to the scenario of WSNs.