Universal AI-Based Intrusion Detection System for In-Vehicle Networks

Although modern vehicles use common communication protocols such as Controller Area Network (CAN), each manufacturer employs different message identifiers, structures, and communication patterns. Consequently, conventional intrusion detection systems often require vehicle-specific training and configuration, limiting their scalability.

To overcome this limitation, LISA developed a Universal Intrusion Detection System (UIDS) designed to identify cyberattacks across different vehicle manufacturers without requiring an entirely separate detection system for each model.

Using real vehicles from BMW, Hyundai, Kia, and Tesla, our research collected and analyzed heterogeneous in-vehicle network traffic to develop a unified AI-based detection approach.

By leveraging machine learning, statistical analysis, and traffic feature extraction, the project aims to identify common characteristics of malicious network behavior regardless of manufacturer-specific CAN messages.

Moving from vehicle-specific intrusion detection toward a universal, scalable cybersecurity framework for connected vehicles.