PoisonSpot
PublicInspired by dynamic taint tracking, PoisonSpot uses fine-grained training provenance tracker that: (1) tags & traces the impact of every single training sample on model updates, (2) probabilistically scores suspect samples based on their linage of impact on model weights, and (3) separates the clean from the poisonous before retraining a model.
adversarial-machine-learningaisecbackdoor-attacksbackdoor-defensebackdoor-detectiondeep-learning-securitymachine-learning-securitypoisoning-attackpoisoning-defensesprovenance-tracking
Creat：2024-04-03T03:03:50
Update：2025-07-14T21:25:23
https://github.com/um-dsp/PoisonSpot
0
Stars
0
Stars Increase