Timezone: »

Reward Poisoning in Reinforcement Learning: Attacks Against Unknown Learners in Unknown Environments
Amin Rakhsha · Xuezhou Zhang · Jerry Zhu · Adish Singla
Event URL: https://openreview.net/forum?id=WCj2fklBR6Y »

We study black-box reward poisoning attacks against reinforcement learning (RL), in which an adversary aims to manipulate the rewards to mislead a sequence of RL agents with unknown algorithms to learn a nefarious policy in an environment unknown to the adversary a priori. That is, our attack makes minimum assumptions on the prior knowledge of the adversary: it has no initial knowledge of the environment or the learner, and neither does it observe the learner's internal mechanism except for its performed actions. We design a novel black-box attack, U2, that can provably achieve a near-matching performance to the state-of-the-art white-box attack, demonstrating the feasibility of reward poisoning even in the most challenging black-box setting.

Author Information

Amin Rakhsha (University of Toronto)
Xuezhou Zhang (Princeton)
Jerry Zhu (University of Wisconsin-Madison)
Adish Singla (MPI-SWS)

More from the Same Authors