KPCA-IPSO-OCSVM方法在工业控制系统入侵检测中的应用

KPCA-IPSO-OCSVM algorithm and its application in intrusion detection of industrial control system

为提高工业控制系统入侵检测的准确性,面向Modbus TCP协议的工业控制系统提出一种基于KPCA-IPSO-OCSVM算法的入侵检测方法。首先采用核主成分分析(kernel principal component analysis,KPCA)方法对强非线性、高复杂度和高维度的工业数据进行特征提取,消除冗余特征,降低数据维度;然后采用免疫粒子群(immune particle swarm optimization,IPSO)优化算法单类支持向量机(one class support vector machine,OCSVM)构建更准确的入侵检测模型。在实验室建立仿真环境,模拟工业控制系统的运行场景,实验结果表明,所提出方法可以精确甄别异常行为,提升入侵检测的准确性和工业控制系统的安全性。

To improve accuracy of industrial control system intrusion detection,an industrial intrusion detection method based on KPCA-IPSO-OCSVM algorithm is proposed for the industrial control system with Modbus TCP protocol.Firstly,the kernel principal component analysis(KPCA)algorithm is used to eliminate redundant features and reduce data dimensions by extracting features of strong nonlinear,high complexity and high dimensional industrial data.Secondly,by using the immune particle swarm optimization(IPSO)algorithm to optimize one class support vector machine(OCSVM),the accuracy of intrusion detection model is furtherly improved.A simulation environment is established in the laboratory to simulate the operating scenarios of industrial control system.Experimental results confirm that the proposed method can accurately identify abnormal behaviors and improve the intrusion detection accuracy and industrial control system security.