In-Vehicle Intrusion Detection based on Machine Learning
O. Saber, Tomader Mazri · 2023
The in-vehicle network enables communication between the different vehicle components and systems. It uses specific protocols and bus systems, such as the Controller Area Network (CAN) that provides communication between invehicle parts simultaneously. As the CAN bus plays an essential role in the in-vehicular communication, it is vulnerable to various types of attack, leading to network disconnection and other damage. This paper proposes an intrusion detection system (IDS) based on machine learning that is able to detect and classify CAN messages as DoS, Fuzzy, Spoofing attacks, or normal messages. A comparison between six different algorithms of classification was carried out and found that Decision Tree, Logistic Regression, and Random Forest were the best algorithms, providing excellent results in terms of the evaluation metrics (accuracy, precision, recall, f1-score) as well as the execution time. Thus, the proposed IDS provides a high detection rate to effectively identify most of the attacks and a low computational time to improve the system’s efficiency.