Securing the Roads: An Analysis of AI Technologies in the Internet of Vehicles

The Internet of Vehicles (IoV) is a relatively new technology that enables smart, connected transportation networks but is increasingly susceptible to internet-based attacks such as spoofing and Denial-of-Service (DoS), which jeopardize user security and system integrity. Conventional Artificial Intelligence (AI) or Machine Learning (ML) models have been used in IoV security, but their limited flexibility and inability to handle complex, evolving attack patterns limit their usefulness in real-world IoV scenarios. This work examines the application of AI/ML models, such as Long Short-Term Memory (LSTM), Deep Reinforcement Learning (DRL), Random Forest (RF), and Extreme Gradient Boosting (XGBoost), to improve cybersecurity in IoV through both proactive and reactive defense strategies. An extensive literature review of current ML-based IoV security solutions is conducted, and AI/ML-based detection models are designed using CICIoV2024, a large-scale dataset obtained from experiments on the Electronic Control Units (ECUs) of a 2019 Ford car. Based on extensive experiments and analysis, we found that the LSTM model can outperform other AI/ML models in identifying spoofing and DoS attacks, thereby enhancing the security of next-generation IoV. Our work can pave the way for understanding the role of the latest AI/ML techniques in enhancing IoV security and the implementation challenges they pose in real-world scenarios.