Securing Autonomous Vehicles: An In-Depth Review of Cyber Attacks and Anomaly Detection Challenges

Autonomous Vehicle (AV) initiatives have rapidly grown in recent years, significantly impacting daily life and enhancing transportation safety and efficiency. Autonomous driving technology promises a future with fully self-driving vehicles while presenting new challenges in safety assurance. In this review, the evolution, obstacles, and methodologies of the statistical approaches for defining cyber threats and discovering anomalies in AVs, specifically under negative conditions and different datasets, are discussed. More critically, this survey assesses the strengths and weaknesses of these methods for their current and future directions. Discussing anomaly detection in AVs under adverse conditions through Federated Learning (FL) and Deep Learning (DL) techniques enhances threat detection capabilities. Additionally, the review explores security vulnerabilities of intra-vehicle and inter-vehicle communication systems concerning various sensors and perception systems, and examines possible attacks on AV software and hardware, emphasising their effects. In addition, the research proposes defensive schemes, founded on statistical methods, DL, optimisations, FL, and blockchain, for strengthening AVs' security. The review aims to improve AVs' resilience against cyberattacks in reconstructing the weaknesses in sensor and perception systems, thereby resulting in the growth and safety of self-driving technology. Furthermore, the review covers anomaly detection in various scenarios, examining advancements in methodologies for better detection performance. Performance evaluations using publicly available datasets are thoroughly analysed, offering a comprehensive overview of current research trends and suggesting pathways for future improvements in AV detection technology.