English

Policy Teaching via Environment Poisoning: Training-time Adversarial Attacks against Reinforcement Learning

Machine Learning 2020-08-20 v2 Artificial Intelligence Cryptography and Security Machine Learning

Abstract

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 average reward in undiscounted infinite-horizon problem settings. The attacker can manipulate the rewards or 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 \emph{optimal stealthy attack} for different measures of attack cost. We provide sufficient technical conditions under which the attack is feasible and provide lower/upper bounds on the attack cost. We instantiate our attacks in two settings: (i) an \emph{offline} setting where the agent is doing planning in the poisoned environment, and (ii) an \emph{online} setting where the agent is learning a policy using a regret-minimization framework 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.

Keywords

Cite

@article{arxiv.2003.12909,
  title  = {Policy Teaching via Environment Poisoning: Training-time Adversarial Attacks against Reinforcement Learning},
  author = {Amin Rakhsha and Goran Radanovic and Rati Devidze and Xiaojin Zhu and Adish Singla},
  journal= {arXiv preprint arXiv:2003.12909},
  year   = {2020}
}

Comments

ICML 2020

R2 v1 2026-06-23T14:30:33.221Z