Research map / Adversarial robustness and security / Attacks on text and graphs
CMUA-Watermark: A Cross-Model Universal Adversarial Watermark for Combating Deepfakes
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Combine two-level perturbation fusion with automatic step-size tuning (ablation) | CMUA-Watermark · CelebA, StarGAN, AGGAN, AttGAN, HiSD | SR_mask: 84.5% -> 100.0%; 99.7% -> 99.9%; 1.5% -> 87.1%; 0.1% -> 99.9% | Helped |
| Add perturbation fusion without automatic step-size tuning (ablation) | CMUA-Watermark · CelebA / StarGAN | SR_mask: 84.5% -> 27% | Hurt |
| Add perturbation fusion without automatic step-size tuning (ablation) | CMUA-Watermark · CelebA / AttGAN | SR_mask: 1.5% -> 82.5% | Helped |
| Add automatic step-size tuning without perturbation fusion (ablation) | CMUA-Watermark · CelebA, StarGAN, AGGAN, AttGAN, HiSD | SR_mask: 84.5% -> 99.0%; 99.7% -> 100.0%; 1.5% -> 20.4%; 0.1% -> 5.0% | Helped |
| Replace PGD with MIM as the base attack in CMUA (ablation) | CMUA-Watermark · CelebA / StarGAN, AGGAN, AttGAN, HiSD | SR_mask: 0.999 -> 1.000; 0.999 -> 0.999; 0.820 -> 0.871; 0.999 -> 0.999 | No clear effect |
| Increase the perturbation bound epsilon (ablation) | CMUA-Watermark · Deepfake-generated facial images | image distortion | Helped |
Closest papers by meaning
- Watermark-embedded Adversarial Examples for Copyright Protection against Diffusion Models (CVPR 2024 · similarity 0.86)
- Robustness of AI-Image Detectors: Fundamental Limits and Practical Attacks (ICLR 2024 · similarity 0.86)
- Once and for All: Universal Transferable Adversarial Perturbation against Deep Hashing-Based Facial Image Retrieval (AAAI 2024 · similarity 0.86)
- TraceEvader: Making DeepFakes More Untraceable via Evading the Forgery Model Attribution (AAAI 2024 · similarity 0.86)
- Lightweight-Mark: Rethinking Deep Learning-Based Watermarking (ICML 2025 · similarity 0.85)
- Guard Me If You Know Me: Protecting Specific Face-Identity from Deepfakes (NeurIPS 2025 · similarity 0.84)
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.