A switching gradient-based algorithm jointly optimizes a backdoor policy and finite-memory observation-based trigger for stealthy attacks in MDPs under partial observations.
TrojDRL: Evaluation of Backdoor Attacks on Deep Reinforcement Learning,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.SY 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Planning Stealthy Backdoor Attacks in MDPs with Observation-Based Triggers
A switching gradient-based algorithm jointly optimizes a backdoor policy and finite-memory observation-based trigger for stealthy attacks in MDPs under partial observations.