AI-Driven Intrusion Detection Systems for Automotive Controller Area Networks

The integration of digital devices in modern vehicles has revolutionized automotive technology, enhancing safety and the overall driving experience. The Controller Area Network (CAN) bus is a central system for managing in-vehicle communication between the electronic control units (ECUs). However, it poses security challenges due to inherent vulnerabilities, lacking encryption and authentication, which, combined with an expanding attack surface, necessitates robust security measures. This thesis investigates Intrusion Detection Systems (IDS) for CAN, a common and cost-effective defense against attacks powered by traditional machine learning and deep learning approaches, aiming to detect and block potential threats. An empirical analysis revealed inconsistencies and a significant performance disparity among state-of-the-art models when tested on various intrusion detection datasets. Notably, the simplicity of the attack strategies contained in established datasets, such as the widely used HCRL Car-Hacking Dataset, has misled researchers, creating a gap between academic research and practical applications. The novel ROAD dataset, with its stealthier and more sophisticated injections, is considered more representative of a real-world scenario, despite having a smaller number of samples and class imbalance. Additionally, on its more realistic attack data, we tested a new hybrid model based on Deep Convolutional Neural Networks (DCNN) and Random Forest, demonstrating its ability to outperform existing models.