Adversarial Attacks/Defence for AI in Autonomous Driving

This presentation offers a comprehensive introduction to the Cybersecurity assessment framework tailored for AI within the realm of "software-defined vehicles" (SDVs). Initially, we outline the anticipated evolution of full-stack IoT architecture solutions, spanning from cloud infrastructure to end-user terminals, and from dedicated platforms to scalable systems. Subsequently, we provide an overview of international regulations and standards governing the ethical use of AI, particularly in the context of SDVs.

Moreover, we address the growing concern of adversarial attacks, including perturbation, which pose significant threats to AI algorithms deployed in autonomous driving scenarios. To mitigate these risks, we propose a robust cybersecurity assessment framework specifically designed for evaluating AI algorithms in automotive applications. Additionally, we introduce a CARLA-based Adversarial Attack Assessment simulation environment, facilitating convenient and faster prototyping for enhanced cybersecurity measures.