BadRL: Sparse Targeted Backdoor Attack against Reinforcement Learning
Proceedings of the AAAI Conference on Artificial Intelligence · 24.03.2024 · 38 (10) · 11687-11694
DOI: 10.1609/aaai.v38i10.29052 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 24.03.2024. Verfasst von Jing Cui und 4 weiteren Autor:innen. Die Arbeit wurde bisher 13-mal zitiert (Crossref).
Abstract (Originalsprache)
Backdoor attacks in reinforcement learning (RL) have previously employed intense attack strategies to ensure attack success. However, these methods suffer from high attack costs and increased detectability. In this work, we propose a novel approach, BadRL, which focuses on conducting highly sparse backdoor poisoning efforts during training and testing while maintaining successful attacks. Our algorithm, BadRL, strategically chooses state observations with high attack values to inject triggers during training and testing, thereby reducing the chances of detection. In contrast to the previous methods that utilize sample-agnostic trigger patterns, BadRL dynamically generates distinct trigger patterns based on targeted state observations, thereby enhancing its effectiveness. Theoretical analysis shows that the targeted backdoor attack is always viable and remains stealthy under specific assumptions. Empirical results on various classic RL tasks illustrate that BadRL can substantially degrade the performance of a victim agent with minimal poisoning efforts (0.003% of total training steps) during training and infrequent attacks during testing. Code is available at: https://github.com/7777777cc/code.
Zitieren
APA
Cui, J., Han, Y., Ma, Y., Jiao, J., & Zhang, J. (2024). BadRL: Sparse Targeted Backdoor Attack against Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 38(10), 11687-11694. https://doi.org/10.1609/aaai.v38i10.29052
Vancouver
Cui J, Han Y, Ma Y, Jiao J, Zhang J. BadRL: Sparse Targeted Backdoor Attack against Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence. 2024;38(10):11687-11694. doi:10.1609/aaai.v38i10.29052
BibTeX
@article{Cui2024,
author = {Cui, Jing and Han, Yufei and Ma, Yuzhe and Jiao, Jianbin and Zhang, Junge},
title = {BadRL: Sparse Targeted Backdoor Attack against Reinforcement Learning},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2024},
volume = {38},
number = {10},
pages = {11687--11694},
doi = {10.1609/aaai.v38i10.29052}
}
EUVIMED