SPA identifies and removes backdoor-watermarked embeddings from EaaS responses by exploiting the constant watermark vector added to triggered text, bypassing verification.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Your Semantic-Independent Watermark is Fragile: A Semantic Perturbation Attack against EaaS Watermark
SPA identifies and removes backdoor-watermarked embeddings from EaaS responses by exploiting the constant watermark vector added to triggered text, bypassing verification.