基于深度学习的车载CAN总线入侵检测方法研究OA
Research on Intrusion Detection Method for Automotive CAN Bus Based on Deep Learning
针对原生缺乏安全机制的控制器局域网(CAN)总线在汽车智能化进程中面临的严峻安全威胁,文章探索了一种高效的入侵检测方法,构建了从数据预处理、特征工程到模型训练的全流程,利用长短期记忆网络(LSTM)结合卷积神经网络(CNN),以实现对异常通信模式的精准识别.在公开数据集上的实验结果表明,本方法对多种常见攻击的检测准确率达到 97.8%,显著优于传统基于规则的检测方法,证实了深度学习模型在车载 CAN 总线入侵检测中的有效性,为解决汽车网络安全问题提供了一条具有高准确性和实用性的技术路径.
To address the severe security threats faced by the inherently insecure Controller Area Network(CAN)bus in the context of vehicle intelligence,this paper explores an efficient intrusion detection method.It constructs a complete process from data preprocessing and feature engineering to model training,and utilizes a combination of Long Short-Term Memory(LSTM)and Convolutional Neural Network(CNN)for accurate identification of abnormal communication patterns.Experimental results on a public dataset show that the proposed method achieves a detection accuracy of 97.8%for various common attacks,significantly outperforming traditional rule-based detection methods.This confirms the effectiveness of the deep learning model for in-vehicle CAN bus intrusion detection,providing a highly accurate and practical technical solution to automotive cybersecurity challenges.
徐进;樊龙;陆续
中国航空工业集团公司西安航空计算技术研究所,陕西 西安 710068中国航空工业集团公司西安航空计算技术研究所,陕西 西安 710068中国航空工业集团公司西安航空计算技术研究所,陕西 西安 710068
信息技术与安全科学
CAN总线入侵检测注意力机制深度学习
CAN busintrusion detectionAttention MechanismDeep Learning
《现代信息科技》 2026 (8)
173-177,183,6
评论