Intrusion Detection Model for In-vehicle CAN Bus Based on TPE-LightGBM Algorithm

The Controller Area Network (CAN) bus is the most widely used in-vehicle communication protocol, which is susceptible to a variety of cyber-attacks due to its lack of security mechanisms such as message authentication and encryption. In order to effectively improve the security communication guarantee of in-vehicle network, this paper proposes a lightweight CAN bus intrusion detection model based on Bayesian optimization LightGBM algorithm, and combined with the actual hacking dataset, the flooding attack, spoofing attack, replaying attack, and fuzzy attack in the CAN bus network are detected by binary classification and multi-classification of attack messages respectively. The experimental results show that the detection precision in multi-classification experiment is 99.19%.