LCBT is a trajectory-only tree-search attack that claims to steer continuous-action RL agents to target policies with sublinear attack cost, but the proof of the claim has a serious importance-sampling flaw.
Vulnerability of deep reinforcement learning to policy induction attacks
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Provably Efficient Action-Manipulation Attack Against Continuous Reinforcement Learning
LCBT is a trajectory-only tree-search attack that claims to steer continuous-action RL agents to target policies with sublinear attack cost, but the proof of the claim has a serious importance-sampling flaw.