Research map / Adversarial robustness and security / Attacks on text and graphs
Uncovering and Mitigating Destructive Multi-Embedding Attacks in Deepfake Proactive Forensics
Lixin Jia, Haiyang Sun, Zhiqing Guo, Yunfeng Diao, Dan Ma, Gaobo Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Apply a second watermark embedding (MEA) | existing proactive forensics methods · CelebA-HQ | BER (%): 0.50 -> 38.17 | Hurt |
| Add AIS fine-tuning (simulate MEA with resilience loss) (ablation) | SepMark · CelebA-HQ, 2 embeddings | BER (%): 48.6223 -> 0.0465 | Improved |
| Add AIS fine-tuning (ablation) | HiDDeN · CelebA-HQ, 2 embeddings | BER (%): 49.5370 -> 2.0679 | Improved |
| Add AIS fine-tuning (ablation) | EditGuard · CelebA-HQ, 5 embeddings | BER (%): 34.7820 -> 2.2408 | Improved |
| Raise AIS loss weight lambda_A from 0 to 10 (ablation) | LampMark · CelebA-HQ, 5 embeddings | BER (%): 18.60 -> 0.22 | Improved |
| Raise AIS loss weight lambda_A from 0 to 10 (ablation) | LampMark · CelebA-HQ, 5 embeddings | PSNR (dB): 17.93 -> 20.32 | Hurt |
| Add AIS fine-tuning (ablation) | HiDDeN · CelebA-HQ, 1 embedding | PSNR (dB): 25.2963 -> 22.0898 | Hurt |
| Add AIS fine-tuning (ablation) | LampMark · CelebA-HQ, 1 embedding | PSNR (dB): 27.0208 -> 28.6729 | Improved |
| Add AIS fine-tuning, evaluated cross-dataset (ablation) | SepMark · LFW, 2 embeddings | BER (%): 49.27 -> 20.16 | Improved |
| Add AIS fine-tuning against cross-model MEA (ablation) | EditGuard · Cross-model MEA, SepMark attacker n=1 | BER (%): 50.51% -> 23.17% | Improved |
| Add AIS fine-tuning against cross-model MEA (ablation) | LampMark defending vs SepMark attacker · Cross-model MEA | BER (%): 11.05% -> 10.04% | Mixed |
Closest papers by meaning
- StableGuard: Towards Unified Copyright Protection and Tamper Localization in Latent Diffusion Models (NeurIPS 2025 · similarity 0.86)
- Robustness of AI-Image Detectors: Fundamental Limits and Practical Attacks (ICLR 2024 · similarity 0.86)
- ROBIN: Robust and Invisible Watermarks for Diffusion Models with Adversarial Optimization (NeurIPS 2024 · similarity 0.85)
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.85)
- Lightweight-Mark: Rethinking Deep Learning-Based Watermarking (ICML 2025 · similarity 0.85)
- Transferable Black-Box One-Shot Forging of Watermarks via Image Preference Models (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
