CADS: Cyber-Attack Detection System Using Decision Tree for Automotive Cyber-Physical Systems

Modern vehicles are equipped with a large number of electronic control units (ECUs) to provide more convenient and safe services for drivers. These ECUs are required to share the various control data for driving using the CAN (Controller Area Network) protocol that is one of the communication protocols for in-vehicle networks. Recently, as vehicles become capable of real-time, two-way communication with various elements on the road, modern vehicles faces a wide range of security issues. The in-vehicle networks are already suffered from the cyber-attacks exploiting newly exposed security vulnerabilities. However, it cannot guarantee the safety and security of the vehicles from as the CAN is designed without considering security. Therefore, this paper proposes a machine learning-based cyber-attack detection system (CADS) in CAN to ensure driving safety. This CADS learns the data pattern of the CAN frames generated while the driving and is able to distinguish between regular messages and cyberattack messages. The performance of the fully trained detection model of the CADS is evaluated using a vast number of testing dataset. The results show that our CADS in terms of accuracy, precision, recall, and F1 score has 100% of the detection rate for regular messages, replay attacks, brute-force attacks except for modification messages.