Policy Teaching via Environment Poisoning: Training-time Adversarial Attacks against Reinforcement Learning
Rakhsha, Amin, Radanovic, Goran, Devidze, Rati, Zhu, Xiaojin, Singla, Adish
–arXiv.org Artificial Intelligence
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.
arXiv.org Artificial Intelligence
Mar-28-2020
- Country:
- North America > United States
- Wisconsin > Dane County > Madison (0.04)
- Asia > Middle East
- Jordan (0.04)
- North America > United States
- Genre:
- Research Report > New Finding (0.68)
- Industry:
- Information Technology > Security & Privacy (1.00)
- Technology: