Data-Driven Solutions for Next-Generation Automotive Cybersecurity

As automotive systems increasingly rely on Ethernet for communication, detecting cyber-attacks on Automotive Ethernet networks becomes crucial to ensure vehicle safety and security. This paper provides an overview of the challenges and methodologies associated with detecting such cyber-attacks. With the complexity of in-vehicle networks growing and the adoption of connected and autonomous vehicle technologies, the risk of cyber-attacks targeting critical vehicle functions has risen. This article explores various cyber-attack scenarios specific to Automotive Ethernet networks, including Frame Injection Attack (FIA), Time Synchronization Attack (TSA), Controller Area Network- Denial of Service (CAN-DoS) attack, CAN Replay (CAN-R) attack, and MAC flooding attack. It proposes a data-driven intrusion detection methodology for automotive cybersecurity, leveraging machine learning and deep learning classifiers. The methodology involves training Machine Learning (ML) and Deep Learning (DL) classifier models using the TOW-IDS( Automotive Ethernet Intrusion Dataset), and evaluating their performance using metrics such as accuracy, precision, and F1 scores. Experimental results demonstrate that the Artificial Neural Network (ANN) classifier achieves high detection performance across five different attack types and is well-suited for automotive-grade devices for real-time anomaly detection. The detection algorithm is implemented in the Python 3.10 platform, and its results are explored.