Research map / Adversarial robustness and security / Black-box and transfer attacks
Untraceable DeepFakes via Traceable Fingerprint Elimination
Jiewei Lai, Lan Zhang, Chen Tang, Pengcheng Sun, Xinming Wang, YUNHAO WANG
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 |
|---|---|---|---|
| add GBMS smoothing (ablation) | multiplicative attack · Attribution evasion (avg over 6 AMs) | ASR (%): 89.82 -> 97.08 | Improved |
| add spectral loss (ablation) | multiplicative attack · Attribution evasion (avg over 6 AMs) | ASR (%): 80.31 -> 97.08 | Improved |
| add spatial loss (ablation) | multiplicative attack · Attribution evasion (avg over 6 AMs) | ASR (%): 89.82 -> 97.08 | Improved |
| add transformation unit to data synthesis (ablation) | multiplicative attack · Attribution evasion (avg over 6 AMs) | ASR (%): 93.80 -> 97.08 | Improved |
| add sampling unit to data synthesis (ablation) | multiplicative attack · Attribution evasion (avg over 6 AMs) | ASR (%): 95.32 -> 97.08 | Improved |
| defender applies adversarial training to AM | TraceEvader (additive attack) · DNA-Det | ASR (%): 98.28 -> 25.10 | Hurt |
| defender adversarial training with the attack's own images | Ours (multiplicative attack) · DNA-Det | ASR (%): 100.00 under white-box defense | No clear effect |
| defender trains network to invert adversarial images | Ours (multiplicative attack) · DnCNN and AutoEncoder inversion | ASR (%): 97.68 and 99.97 after inversion | No clear effect |
| multiplicative instead of additive attack (Ours vs TraceEvader) | attribution evasion · Attribution evasion (avg over 6 AMs) | ASR (%): 87.11 -> 97.08 | Improved |
| multiplicative instead of additive attack (Ours vs TraceEvader) | attribution evasion · Image fidelity | SSIM: 0.995 -> 0.963 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TraceEvader — TraceEvader: Making DeepFakes More Untraceable via Evading the Forgery Model Attribution (AAAI 2024) · better on 3 of 5 shared results
- DiffAttack — DiffAttack: Evasion Attacks Against Diffusion-Based Adversarial Purification (NeurIPS 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TraceEvader: Making DeepFakes More Untraceable via Evading the Forgery Model Attribution (AAAI 2024 · similarity 0.90)
- Artificial Fingerprinting for Generative Models: Rooting Deepfake Attribution in Training Data (ICCV 2021 · similarity 0.87)
- Where the Devil Hides: Deepfake Detectors Can No Longer Be Trusted (CVPR 2025 · similarity 0.86)
- Unlearn and Burn: Adversarial Machine Unlearning Requests Destroy Model Accuracy (ICLR 2025 · similarity 0.85)
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.84)
- Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing (AAAI 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.
