基于特征信息熵与支持向量机的智能网联汽车CAN总线异常检测技术  

Anomaly Detection Technology of Intelligent Networked Vehicle CAN Bus Based on Feature Information Entropy and Support Vector Machine

作　　者：陈宁[1] Chen Ning(Zhejiang Institute of Mechanical&Electrical Engineer,Hangzhou,China)

机构地区：[1]浙江机电职业技术学院,浙江杭州

出　　处：《科学技术创新》2024年第7期63-66,共4页Scientific and Technological Innovation

基　　金：浙江省科技厅公益基金项目《基于边缘计算的车路协同智能路侧单元关键技术研究(LGG22F020031)》。

摘　　要：本文结合CNA报文的结构特点,探究了基于特征、信息熵的异常检测技术和基于支持向量机的异常检测技术。基于特征、信息熵的异常检测技术,将CAN ID作为特征,统计包含该特征的所有报文并计算信息熵。根据信息熵确立阈值标准,对比CAN总线报文的熵值是否在阈值范围内,从而检测是否存在异常。仿真结果表明,在报文数量较少的情况下,该技术的异常检测率可以达到100%。基于支持向量机的异常检测技术,将异常报文预处理后输入到支持向量机中训练,得到异常检测指标。利用该指标与CAN总线报文进行对比,从而检测是否存在异常。实验结果表明,该技术对多种CNA报文的异常检测率在90%以上。Based on the structural characteristics of CNA messages,this paper explores the anomaly detection technology based on feature and information entropy and the anomaly detection technology based on support vector machine.The anomaly detection technology based on feature and information entropy takes CAN ID as a feature,collects statistics on all packets containing the feature,and calculates the information entropy.Establish a threshold standard based on information entropy,and compare the entropy of CAN bus packets to check whether the entropy is within the threshold range.Simulation results show that the anomaly detection rate of this technology can reach 100% when the number of packets is small.Based on the support vector machine(SVM) anomaly detection technology,the abnormal message is preprocessed and input into SVM for training,and the anomaly detection index is obtained.The indicator is compared with the CAN bus message to detect whether there is an exception.The experimental results show that the anomaly detection rate of various CNA messages is more than 90%.

关 键 词：信息熵 支持向量机 CAN总线 异常检测 

分 类 号：U463.6[机械工程—车辆工程] TN925.93[交通运输工程—载运工具运用工程]