CoRL

Exposing Vulnerabilities in RL: A Novel Stealthy Backdoor Attack through Reward Poisoning

揭示强化学习中的漏洞：通过奖励污染的新型隐秘后门攻击

Bokang Zhang, Chaojun Lu, Jianhui Li, Junfeng Wu

AI总结 本文提出通过奖励污染对强化学习代理进行隐秘后门攻击，展示了在不同环境中攻击的有效性和隐蔽性。

Comments Workshop on Safe and Robust Robot Learning for Operation in the Real World at CoRL 2025