BACKDOORL: Backdoor Attack against Competitive Reinforcement Learning

Lun Wang, Zaynah Javed, Xian Wu, Wenbo Guo, Xinyu Xing, Dawn Song

Research output: Chapter in Book/Report/Conference proceedingConference contribution

21 Scopus citations

Abstract

Recent research has confirmed the feasibility of backdoor attacks in deep reinforcement learning (RL) systems. However, the existing attacks require the ability to arbitrarily modify an agent's observation, constraining the application scope to simple RL systems such as Atari games. In this paper, we migrate backdoor attacks to more complex RL systems involving multiple agents and explore the possibility of triggering the backdoor without directly manipulating the agent's observation. As a proof of concept, we demonstrate that an adversary agent can trigger the backdoor of the victim agent with its own action in two-player competitive RL systems. We prototype and evaluate BACKDOORL in four competitive environments. The results show that when the backdoor is activated, the winning rate of the victim drops by 17% to 37% compared to when not activated. The videos are hosted at https://github.com/wanglun1996/multi agent rl backdoor videos.

Original languageEnglish (US)
Title of host publicationProceedings of the 30th International Joint Conference on Artificial Intelligence, IJCAI 2021
EditorsZhi-Hua Zhou
PublisherInternational Joint Conferences on Artificial Intelligence
Pages3699-3705
Number of pages7
ISBN (Electronic)9780999241196
StatePublished - 2021
Event30th International Joint Conference on Artificial Intelligence, IJCAI 2021 - Virtual, Online, Canada
Duration: Aug 19 2021Aug 27 2021

Publication series

NameIJCAI International Joint Conference on Artificial Intelligence
ISSN (Print)1045-0823

Conference

Conference30th International Joint Conference on Artificial Intelligence, IJCAI 2021
Country/TerritoryCanada
CityVirtual, Online
Period8/19/218/27/21

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'BACKDOORL: Backdoor Attack against Competitive Reinforcement Learning'. Together they form a unique fingerprint.

Cite this