基于梯度提升决策树算法的控制器局域网总线异常检测  

A control area network bus anomaly detection based on gradient boosting decision tree algorithm

作　　者：王杰 何雨桐 莫秀良 WANG Jie;HE Yutong;MO Xiuliang(School of Computer Science and Engineering,Tianjin University of Technology,Tianjin 300384,China)

机构地区：[1]天津理工大学计算机科学与工程学院,天津300384

出　　处：《天津理工大学学报》2022年第3期26-31,共6页Journal of Tianjin University of Technology

基　　金：天津市科技特派员项目(18JCTPJC51000)。

摘　　要：车辆的网联化和自动化已成为汽车发展的必然趋势,控制器域网(controller area network,CAN)是汽车总线网络中最常用的协议,但由于它没有足够的安全功能,如消息加密和发送者身份验证,无法保护车辆网络免受攻击,因此有必要制订适当的对策来保证CAN的安全。由此提出一种针对数据域的CAN总线异常检测模型,使用真实车辆构建的数据集进行试验研究,在没有总线报文标识功能的情况下,根据每个标识符(identifier,ID)的数据域分布特点,实现了基于梯度提升决策树(gradient boosting decision tree,GBDT)算法的混合模型。模拟攻击试验结果表明,该异常检测模型在抵御篡改数据域消息方面具有较高的准确率。The networking and automation of vehicles are an inevitable trend in the development of automobiles. The controller area network(CAN)is the most commonly used protocol in automotive bus networks,however,it lacks sufficient security features,such as message encryption and sender identity verification. Vehicle network is vulnerable to malicious attacks and it is necessary to develop countermeasures to ensure the security of controller area network. Therefore,a controller area network of the bus anomaly detection model for message data domain is proposed,an experimental study was conducted using the dataset constructed by real vehicles,and a hybrid model based on gradient boosting decision tree(GBDT)algorithm was implemented according to the distribution characteristics of the data domain of each identifier(ID)in the absence of the bus message identification function. The experimental results through simulating the attacks show that the anomaly detection model has a promising accuracy rate in resisting tampered data domain messages.

关 键 词：网联汽车 控制器局域网总线 信息安全 控制器局域网标识符 异常检测 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]