Abstract
Recently, autonomous and connected vehicles have gained popularity, revolutionizing consumer mobility. On the other hand, they are also becoming new targets exposing new attack vectors and vulnerabilities that may lead to critical consequences. In this paper, we propose VNGuard, an intrusion detection system for two critical in-vehicle networks (IVNs), namely, the Local Interconnect Network (LIN) and the Automotive Ethernet (AE). In the proposed system, LIN messages and AE network packets are converted into images, and then a state-of-theart deep convolutional neural networks (DCNN) model is applied to not only detect anomalous traffic, but also to classify types of attacks. Our experimental results showed that the VNGuard achieves more than 96% detection accuracy for LIN and 99% attack classification accuracy for AE. In addition, the VNGuard is able to perform the intrusion detection within 3 ms for LIN and 4 ms for AE significantly within the latency constraint required by the autonomous and connected vehicles to achieve human-level safety.