Research map / Adversarial robustness and security / Attacks on text and graphs
Active Adversarial Noise Suppression for Image Forgery Localization
R. C. Peng, Shunquan Tan, Xianbo Mo, Alex Chichung Kot, Jiwu Huang
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 ANSM defensive perturbation preprocessing | CoDE · Columbia under BIM attack | pixel-level F1: 0.005 -> 0.827 | Improved |
| add ANSM defensive perturbation preprocessing | MVSS-Net · IMD20 under BIM attack | pixel-level F1: 0.038 -> 0.250 | Improved |
| add ANSM defensive perturbation preprocessing | CoDE · six attacks average RP | RP (%): 24.3% -> 99.7% under FGSM; 3.3% -> 95.9% under BIM | Improved |
| add ANSM preprocessing on original (unattacked) images | IF-OSN · Columbia original | pixel-level F1: 0.706 -> 0.693 | Hurt |
| add ANSM preprocessing on original images | MVSS-Net · Columbia original | pixel-level F1: 0.677 -> 0.611 | Hurt |
| use forgery-relevant feature alignment instead of pixel-wise reconstruction (ablation) | ANSM IF-OSN · Ablation FGSM (CASIAv1, IMD20, MISD test) | RP under FGSM (%): 49.3 -> 89.4 | Improved |
| use FFA on CoDE instead of PwR (ablation) | ANSM CoDE · Ablation FGSM (CASIAv1, IMD20, MISD test) | RP under FGSM (%): 50.6 -> 91.1 | Improved |
| align middle-layer features instead of shallow-layer features (ablation) | ANSM IF-OSN · Ablation FGSM (CASIAv1, IMD20, MISD test) | RP under FGSM (%): 49.4 -> 89.4 | Improved |
| align middle-layer instead of topmost-layer features (ablation) | ANSM IF-OSN · Ablation FGSM (CASIAv1, IMD20, MISD test) | RP on original images (%): 85.4 -> 87.5 | Improved |
| add MgR with M_o mask after FFA (ablation) | ANSM IF-OSN · Ablation FGSM (CASIAv1, IMD20, MISD test) | RP under FGSM (%): 89.4 -> 99.2 | Improved |
| add FFA to MgR-only training (ablation) | ANSM IF-OSN · Ablation FGSM (CASIAv1, IMD20, MISD test) | pixel-level F1 under FGSM: 0.216 -> 0.503 | Improved |
| use M_o instead of M_gt as supervised mask in MgR (ablation) | ANSM IF-OSN · Ablation FGSM (CASIAv1, IMD20, MISD test) | pixel-level F1 on original: 0.423 -> 0.477 | Improved |
Closest papers by meaning
- Learning Discriminative Noise Guidance for Image Forgery Detection and Localization (AAAI 2024 · similarity 0.89)
- Detecting Adversarial Data Using Perturbation Forgery (CVPR 2025 · similarity 0.89)
- Towards Better Robustness Against Natural Corruptions in Document Tampering Localization (AAAI 2025 · similarity 0.87)
- PAD: Patch-Agnostic Defense against Adversarial Patch Attacks (CVPR 2024 · similarity 0.87)
- MGQFormer: Mask-Guided Query-Based Transformer for Image Manipulation Localization (AAAI 2024 · similarity 0.86)
- Adversarial Robust Safeguard for Evading Deep Facial Manipulation (AAAI 2024 · similarity 0.86)
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.
