Policy Teaching in Reinforcement Learning via Environment Poisoning Attacks
This work highlights a significant security threat for reinforcement learning agents in practical applications, showing how an attacker can manipulate the learning process.
This paper investigates a security threat in reinforcement learning where an attacker manipulates the training environment to force an agent to adopt a specific target policy. The authors propose an optimization framework to find stealthy attacks, demonstrating that an attacker can easily succeed in teaching any target policy to the victim under mild conditions.
We study a security threat to reinforcement learning where an attacker poisons the learning environment to force the agent into executing a target policy chosen by the attacker. As a victim, we consider RL agents whose objective is to find a policy that maximizes reward in infinite-horizon problem settings. The attacker can manipulate the rewards and the transition dynamics in the learning environment at training-time, and is interested in doing so in a stealthy manner. We propose an optimization framework for finding an optimal stealthy attack for different measures of attack cost. We provide lower/upper bounds on the attack cost, and instantiate our attacks in two settings: (i) an offline setting where the agent is doing planning in the poisoned environment, and (ii) an online setting where the agent is learning a policy with poisoned feedback. Our results show that the attacker can easily succeed in teaching any target policy to the victim under mild conditions and highlight a significant security threat to reinforcement learning agents in practice.