Research map / Image generation / Autoregressive and latent generation
Learning to Immunize Images for Tamper Localization and Self-Recovery
Qichao Ying, Hang Zhou, Zhenxing Qian, Sheng Li, Xinpeng Zhang
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 image protection (immunization) before tampering (ablation) | Imuge+ · Ablation (JPEG QF=70, tampering rate 0.1-0.4) | F1: 0.435 -> 0.965 | Improved |
| add tampering-based data augmentation (ablation) | Imuge+ · Ablation (JPEG QF=70, tampering rate 0.1-0.4) | F1: 0.435 -> 0.965 | Improved |
| add edge as extra input and supervision (ablation) | Imuge+ · Ablation (JPEG QF=70, tampering rate 0.1-0.4) | F1: 0.854 -> 0.965 | Improved |
| add asymmetric batch size (ablation) | Imuge+ · Ablation (JPEG QF=70, tampering rate 0.1-0.4) | F1: 0.913 -> 0.965 | Improved |
| add iterative training (ablation) | Imuge+ · Ablation (JPEG QF=70, tampering rate 0.1-0.4) | F1: failed to train -> 0.965 | Improved |
| use KD-JPEG instead of Diff-JPEG as JPEG simulator (ablation) | Imuge+ · Ablation (JPEG QF=70, tampering rate 0.1-0.4) | F1: 0.933 -> 0.965 | Improved |
| use single IINet instead of two separate U-Nets (ablation) | Imuge+ · Ablation (JPEG QF=70, tampering rate 0.1-0.4) | F1: 0.962 -> 0.965 | Improved |
| use KD-JPEG instead of Diff-JPEG (ablation) | JPEG simulation · JPEG simulation fidelity (QF=10) | PSNR (dB): 28.03 -> 29.64 | Improved |
| use KD-JPEG instead of Diff-JPEG (ablation) | JPEG simulation · JPEG QF classification | accuracy (%): 72.48 -> 95.47 | Improved |
| use Imuge+ instead of Imuge | image immunization · Copy-move, no attack | F1: 0.642 -> 0.903 | Improved |
| use Imuge+ instead of Imuge | image immunization · Inpainting, JPEG | F1: 0.253 -> 0.872 | Improved |
| use Imuge+ instead of best passive method SPAN | tampering localization · Inpainting, JPEG | F1: 0.624 -> 0.872 | Improved |
Closest papers by meaning
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.85)
- Distraction is All You Need: Memory-Efficient Image Immunization against Diffusion-Based Image Editing (CVPR 2024 · similarity 0.85)
- EditGuard: Versatile Image Watermarking for Tamper Localization and Copyright Protection (CVPR 2024 · similarity 0.84)
- StableGuard: Towards Unified Copyright Protection and Tamper Localization in Latent Diffusion Models (NeurIPS 2025 · similarity 0.83)
- DiffForensics: Leveraging Diffusion Prior to Image Forgery Detection and Localization (CVPR 2024 · similarity 0.82)
- Towards Better Robustness Against Natural Corruptions in Document Tampering Localization (AAAI 2025 · similarity 0.82)
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.
