DoS Attack Detection Enhancement in Autonomous Vehicle Systems with Explainable AI

Autonomous Vehicle Systems (AVS) are designed to interact with their environment, including the atmosphere and other smart vehicles, autonomously. These systems possess significant computational and communication capabilities, making them attractive targets for malicious attacks, particularly cyber-attacks. Among the most concerning cyber-attacks is the Denial of Service (DoS) attack, which can potentially cause AVS malfunctions, leading to severe consequences such as accidents, or, in the worst case, loss of human life. To detect and defend against such cyber-attacks, various Intrusion Detection Systems (IDS) have been developed, employing both statistical analysis and artificial intelligence (AI) algorithms. However, these AIbased algorithms are often considered “black boxes” due to their inability to provide explanations for their predictions. In this paper, we propose the integration of Explainable Artificial Intelligence (XAI) to improve the interpretability of AI predictions, thus enhancing the overall prediction performance. For simulation, we utilize the UNSW-NB15 dataset, specifically focusing on its DoS attack features, and demonstrate that incorporating XAI leads to higher prediction accuracy.