Optimal Actuator Attacks on Autonomous Vehicles Using Reinforcement Learning

TL;DR

This paper introduces a reinforcement learning approach to design optimal stealthy attacks on autonomous vehicle actuators and analyzes the limitations of existing secure controllers, validated through extensive simulations.

Contribution

It presents a novel RL-based method for creating optimal actuator attacks on AVs and evaluates the vulnerabilities of current secure control strategies.

Findings

RL-based attacks are highly effective and stealthy

Current secure controllers have significant limitations against these attacks

Simulation results confirm the method's efficiency and impact

Abstract

With the increasing prevalence of autonomous vehicles (AVs), their vulnerability to various types of attacks has grown, presenting significant security challenges. In this paper, we propose a reinforcement learning (RL)-based approach for designing optimal stealthy integrity attacks on AV actuators. We also analyze the limitations of state-of-the-art RL-based secure controllers developed to counter such attacks. Through extensive simulation experiments, we demonstrate the effectiveness and efficiency of our proposed method.