Detection of Vehicle Intrusion Attacks in Controller Area Networks

As modern vehicles become more connected and dependent on the Controller Area Network (CAN) for essential in-vehicle communications, the threat of cyber intrusions has grown substantially. Conventional security solutions fail to offer real-time intrusion detection in high-speed automotive environments because of their computational overhead and inefficiency. This paper proposes a novel Intrusion Detection System (IDS) based on a hybrid deep learning architecture that involves MobileNetV2 for extracting features and FastGRNN for processing sequences. The lightweight design of MobileNetV2 makes it efficient for spatial feature extraction, while temporal dependencies are addressed by FastGRNN, providing real-time detection of anomalies in vehicular networks. Experimental evaluations prove that the new IDS is superior to the traditional CNN + GRU-based models in aspects of computational complexity, accuracy, and inference time. This work advances automotive cybersecurity through the provision of an optimized, low-latency intrusion detection solution that can be deployed in real-time within contemporary intelligent vehicles.