Optimal Attack and Defense for Reinforcement Learning

arXiv:2312.00198v2 Announce Type: replace-cross
Abstract: 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 …

adversarial adversarial attacks agent arxiv attack attacker attacks class cs.cr cs.gt cs.lg defense environment external noise power real study systems victim