Towards Secure Autonomous Driving: Edge AI—Based Cybersecurity for ADAS and the Internet of Autonomous Vehicles (IoAV)

Abstract

This paper systematically investigates the Edge AIbased cybersecurity in ADAS and IoAV. Time is moving forward and cars becoming more autonomous, connected than everReal time trust-worthy Cybersecurity is a vital need to drive safely. The classical cloud-based security alternatives have to deal with limitations such as latency, bandwidth dependence and low responsiveness which are not appropriate for the safety-critical vehicular networks. Edge AI ameliorates this drawback by providing the intelligence at edge for intrusion detection, lightweight authentication secure perception and privacy preserving learning directly on vehicles and base stations on roadside. In this work, we present an overview of major attack surfaces and real-world cybersecurity compromises, followed by the latest defense strategies to prevent them and then argue why it is advantageous to use edge-based strategy rather than centralized. Furthermore, we also provide a comparative study with state-of-the-arts in accuracy, latency, computational cost and application adaptation. Finally, the work highlights open research issues and potential pathways to achieve secure, scalable and trustworthy autonomous transportation systems.