Evaluation of Machine Learning and Deep Learning-based Intrusion Detection Systems in In-vehicle Networks

With the growing complexity and connectivity of Intelligent Connected Vehicles (ICVs), In-vehicle communication plays an important role in today’s driving environment. The Controller Area Network (CAN) is the most used protocol in an in-vehicle network to exchange data between Electronic Control Units (ECUs). Although in-vehicle communications must be regarded as the last line of security defence for ICV, recent reports have indicated vulnerability to threats due to its broadcasting design. In this context, this paper targets to comprehensively review the advantages and limitations of the existing machine learning and deep learning approaches designed for Intrusion Detection Systems (IDS). We evaluate their performance and computational resource requirements and propose potential enhancements. Finally, open challenges and future research directions for in-vehicle communication cyber security are highlighted as observations and recommendations.