TrojanTO implants action-level backdoors into Decision Transformer style offline RL models using 10 trajectories, alternating trigger optimization and model fine-tuning, reaching average attack success 0.719 while preserving benign performance.
We performed an ablation study against a naive single-objective approach, which optimizes only Equa- tion 6 across all data, including poisoned ones
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TrojanTO: Action-Level Backdoor Attacks against Trajectory Optimization Models
TrojanTO implants action-level backdoors into Decision Transformer style offline RL models using 10 trajectories, alternating trigger optimization and model fine-tuning, reaching average attack success 0.719 while preserving benign performance.