EUVIMEDEUVIMED entdecken
Crossref

Optimal Attack and Defense for Reinforcement Learning

Jeremy McMahan, Young Wu, Xiaojin Zhu, Qiaomin Xie

Proceedings of the AAAI Conference on Artificial Intelligence · 24.03.2024 · 38 (13) · 14332-14340

DOI: 10.1609/aaai.v38i13.29346 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 24.03.2024. Verfasst von Jeremy McMahan und 3 weiteren Autor:innen. Die Arbeit wurde bisher 10-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To ensure the usefulness of Reinforcement Learning (RL) in real systems, it is crucial to ensure they are robust to noise and adversarial attacks. In adversarial RL, an external attacker has the power to manipulate the victim agent's interaction with the environment. We study the full class of online manipulation attacks, which include (i) state attacks, (ii) observation attacks (which are a generalization of perceived-state attacks), (iii) action attacks, and (iv) reward attacks. We show the attacker's problem of designing a stealthy attack that maximizes its own expected reward, which often corresponds to minimizing the victim's value, is captured by a Markov Decision Process (MDP) that we call a meta-MDP since it is not the true environment but a higher level environment induced by the attacked interaction. We show that the attacker can derive optimal attacks by planning in polynomial time or learning with polynomial sample complexity using standard RL techniques. We argue that the optimal defense policy for the victim can be computed as the solution to a stochastic Stackelberg game, which can be further simplified into a partially-observable turn-based stochastic game (POTBSG). Neither the attacker nor the victim would benefit from deviating from their respective optimal policies, thus such solutions are truly robust. Although the defense problem is NP-hard, we show that optimal Markovian defenses can be computed (learned) in polynomial time (sample complexity) in many scenarios.

Zitieren

APA

McMahan, J., Wu, Y., Zhu, X., & Xie, Q. (2024). Optimal Attack and Defense for Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 38(13), 14332-14340. https://doi.org/10.1609/aaai.v38i13.29346

Vancouver

McMahan J, Wu Y, Zhu X, Xie Q. Optimal Attack and Defense for Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence. 2024;38(13):14332-14340. doi:10.1609/aaai.v38i13.29346

BibTeX
@article{McMahan2024,
  author = {McMahan, Jeremy and Wu, Young and Zhu, Xiaojin and Xie, Qiaomin},
  title = {Optimal Attack and Defense for Reinforcement Learning},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2024},
  volume = {38},
  number = {13},
  pages = {14332--14340},
  doi = {10.1609/aaai.v38i13.29346}
}
Kostenlos starten →