English

Reinforcement Learning under Threats

Machine Learning 2019-10-28 v2 Artificial Intelligence Cryptography and Security Machine Learning

Abstract

In several reinforcement learning (RL) scenarios, mainly in security settings, there may be adversaries trying to interfere with the reward generating process. In this paper, we introduce Threatened Markov Decision Processes (TMDPs), which provide a framework to support a decision maker against a potential adversary in RL. Furthermore, we propose a level-kk thinking scheme resulting in a new learning framework to deal with TMDPs. After introducing our framework and deriving theoretical results, relevant empirical evidence is given via extensive experiments, showing the benefits of accounting for adversaries while the agent learns.

Keywords

Cite

@article{arxiv.1809.01560,
  title  = {Reinforcement Learning under Threats},
  author = {Victor Gallego and Roi Naveiro and David Rios Insua},
  journal= {arXiv preprint arXiv:1809.01560},
  year   = {2019}
}

Comments

Extends the verson published at the Proceedings of the AAAI Conference on Artificial Intelligence 33, https://www.aaai.org/ojs/index.php/AAAI/article/view/5106

R2 v1 2026-06-23T03:55:16.623Z