摘要
针对日益严峻的车载网络安全问题,提出一种基于卷积神经网络(CNN)的车载网络入侵检测模型。针对车载网络流量数据的特点,首先通过引入GAF编码将车载网络一维时间序列数据转换为二维矩阵,再转换为图片作为CNN网络的输入;为提高CNN网络性能,选择VGG网络作为主干网络,并对VGG网络的损失函数、隐藏层激活函数、分类器、权值初始化等进行优化;最后在通用的DoS攻击数据集和Fuzzy攻击数据集上进行仿真验证。仿真结果表明,所提GAF编码+VGG网络的入侵检测模型可实现汽车CAN总线的DoS攻击和Fuzzy攻击检测,整体检测准确率达到99%以上,且相较于Reduced Inception-Resnet网络入侵检测模型,所构建的入侵检测模型在F1值等指标上更有优势。
Aiming at the increasingly serious problem of vehicle network security,a vehicle network intrusion detection model based on convolutional neural network(CNN)is proposed.According to the characteristics of vehicle network traffic data,firstly,one-dimensional time series data of vehicle network is converted into two-dimensional matrix by introducing GAF coding,and then converted into pictures as the input of CNN network.In order to improve the performance of CNN network,VGG network is selected as the backbone network,and the loss function,hidden layer activation function,classifier and weight initialization of VGG network are optimized.Finally,the simulation verification is carried out on the general DoS attack data set and Fuzzy attack data set.The simulation results show that the proposed intrusion detection model of GAF coding+VGG network CAN realize DoS attack and Fuzzy attack detection of automobile CAN bus,and the overall detection accuracy rate is over 99%.Compared with the reduced incidence-resnet network intrusion detection model,the constructed intrusion detection model has more advantages in F1 value and other indicators.
作者
吴珊
Wu Shan(Department of Automobile,Xianyang Vocational and Technical College,Shaanxi Xianyang,712000,China)
出处
《机械设计与制造工程》
2022年第9期65-69,共5页
Machine Design and Manufacturing Engineering
基金
咸阳职业技术学院科学研究基金项目(2020KJB03)。
