Reward Poisoning in Reinforcement Learning: Attacks Against Unknown Learners in 
Unknown Environments

Rakhsha, Amin; Zhang, Xuezhou; Zhu, Xiaojin; Singla, Adish

Item

ITEM ACTIONSEXPORT

Add to Basket

Local TagsRelease HistoryDetailsSummary

Released

Paper

Reward Poisoning in Reinforcement Learning: Attacks Against Unknown Learners in Unknown Environments

MPS-Authors

/persons/resource/persons216578

Singla, Adish
Group A. Singla, Max Planck Institute for Software Systems, Max Planck Society;

External Resource

No external resources are shared

Fulltext (restricted access)

There are currently no full texts shared for your IP range.

Fulltext (public)

arXiv:2102.08492.pdf
(Preprint), 343KB

Supplementary Material (public)

There is no public supplementary material available

Citation

Rakhsha, A., Zhang, X., Zhu, X., & Singla, A. (2021). Reward Poisoning in Reinforcement Learning: Attacks Against Unknown Learners in Unknown Environments. Retrieved from https://arxiv.org/abs/2102.08492.

Cite as: https://hdl.handle.net/21.11116/0000-0009-F810-D

Abstract

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.