Research map / Adversarial robustness and security / Attacks on text and graphs
Evading DeepFake Detectors via Adversarial Statistical Consistency
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 |
|---|---|---|---|
| Replace conventional additive perturbations with statistical-consistency degradations | DeepFake detector evasion · FF++ transfer from ResNet to DenseNet | ASR: VMIFGSM 55.0% -> StatAttack 81.3% | Helped |
| Add multi-layer degradations with jointly tuned combination weights | StatAttack · FF++ transfer from ResNet to DenseNet | ASR: 81.3% -> 83.1% | Helped |
| Add multi-layer degradations with jointly tuned combination weights | StatAttack · FF++ transfer from ResNet to DCTA | ASR: 18.5% -> 17.3% | Hurt |
| Use adversarial noise only (ablation) | StatAttack · FF++ transfer to EfficientNet | ASR: StatAttack 65.9%; only noise 22.5% | Hurt |
| Use adversarial exposure only (ablation) | StatAttack · FF++ transfer to EfficientNet | ASR: StatAttack 65.9%; only exposure 26.8% | Hurt |
| Use adversarial blur only (ablation) | StatAttack · FF++ transfer to EfficientNet | ASR: StatAttack 65.9%; only blur 10.3% | Hurt |
| Add adversarial noise (ablation) | StatAttack · FF++ transfer to EfficientNet | ASR: 37.5% without noise; 65.9% StatAttack | Helped |
| Add adversarial exposure (ablation) | StatAttack · FF++ transfer to EfficientNet | ASR: 39.5% without exposure; 65.9% StatAttack | Helped |
| Add adversarial blur (ablation) | StatAttack · FF++ transfer to EfficientNet | ASR: 57.1% without blur; 65.9% StatAttack | Helped |
| Add adversarial exposure (ablation) | StatAttack · FF++ transfer to DCTA | ASR: 10.5% without exposure; 35.5% full attack | Helped |
Closest papers by meaning
- DF40: Toward Next-Generation Deepfake Detection (NeurIPS 2024 · similarity 0.90)
- TraceEvader: Making DeepFakes More Untraceable via Evading the Forgery Model Attribution (AAAI 2024 · similarity 0.89)
- Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing (AAAI 2025 · similarity 0.89)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · similarity 0.89)
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.89)
- Where the Devil Hides: Deepfake Detectors Can No Longer Be Trusted (CVPR 2025 · similarity 0.88)
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.