The proposed steganography-based attribution system with CLIP multimodal fusion achieves robust watermarking under distortions and 0.99 AUC-ROC for harm detection, enabling traceable AI content accountability.
C2pa: the world’s first industry standard for content provenance (conference presentation),
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Toward Accountable AI-Generated Content on Social Platforms: Steganographic Attribution and Multimodal Harm Detection
The proposed steganography-based attribution system with CLIP multimodal fusion achieves robust watermarking under distortions and 0.99 AUC-ROC for harm detection, enabling traceable AI content accountability.