Retraining a fine-tuned watermarked neural network on its original clean training data can restore the erased backdoor watermark, without exposing the trigger set.
GPT-4 Technical Report,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Persistence of Backdoor-based Watermarks for Neural Networks: A Comprehensive Evaluation
Retraining a fine-tuned watermarked neural network on its original clean training data can restore the erased backdoor watermark, without exposing the trigger set.