The implementation of electronics in modern vehicles has resulted in an increase in attacks targeting in-vehicle networks; thus, attack detection models have caught the attention of the automotive industry and its researchers. Vehicle network security is an urgent and significant problem because the malfunctioning of vehicles can directly affect human and road safety. The controller area network (CAN), which is used as a de facto standard for in-vehicle networks, does not have sufficient security features, such as message encryption and sender authentication, to protect the network from cyber-attacks. In this paper, we propose an intrusion detection system (IDS) based on a deep convolutional neural network (DCNN) to protect the CAN bus of the vehicle. The DCNN learns the network traffic patterns and detects malicious traffic without hand-designed features. We designed the DCNN model, which was optimized for the data traffic of the CAN bus, to achieve high detection performance while reducing the unnecessary complexity in the architecture of the Inception-ResNet model. We performed an experimental study using the datasets we built with a real vehicle to evaluate our detection system. The experimental results demonstrate that the proposed IDS has significantly low false negative rates and error rates when compared to the conventional machine-learning algorithms.

在现代车辆中使用电子设备导致针对车载网络的攻击有所增加；因此，攻击检测模型引起了汽车工业及其研究人员的关注。车辆网络安全是一个紧迫而重大的问题，因为车辆的故障会直接影响人身和道路安全。用作局域网的事实上的标准的控制器局域网（CAN）没有足够的安全功能（例如消息加密和发送方身份验证）来保护网络免受网络攻击。在本文中，我们提出了一种基于深度卷积神经网络（DCNN）的入侵检测系统（IDS），以保护车辆的CAN总线。DCNN无需手动设计的功能即可了解网络流量模式并检测恶意流量。我们设计了DCNN模型，该模型针对CAN总线的数据流量进行了优化，以实现较高的检测性能，同时减少Inception-ResNet模型的体系结构中不必要的复杂性。我们使用由真实车辆构建的数据集进行了实验研究，以评估我们的检测系统，实验结果表明，与传统的机器学习算法相比，所提出的IDS的误报率和误码率都非常低。