Research map / Image generation / Fast diffusion sampling
Enhancing Membership Inference Attacks on Diffusion Models from a Frequency-Domain Perspective
Puwei Lian, Yujun Cai, Songze Li, Bingkun BAO
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 high-frequency filter | Naive attack · MS-COCO (fine-tuned SD) | AUC: 87.85 -> 98.32 | Improved |
| add high-frequency filter | PIA attack · MS-COCO (fine-tuned SD) | TPR@1%FPR: 5.20 -> 16.59 | Improved |
| add high-frequency filter | SecMI attack · Tiny-IN (DDIM) | TPR@1%FPR: 13.96 -> 25.79 | Improved |
| add high-frequency filter | Naive attack · Flickr (fine-tuned SD) | TPR@1%FPR: 16.59 -> 67.60 | Improved |
| add high-frequency filter | Naive attack · MS-COCO (weak overfitting) | AUC: 63.51 -> 75.43 | Improved |
| add high-frequency filter | CLID attack · MS-COCO (typical) | AUC: 99.43 -> 99.51 | No clear effect |
| add high-frequency filter | CLID attack · MS-COCO (weak overfitting) | AUC: 93.32 -> 95.41 | Improved |
| add high-frequency filter | attacks on pre-trained Stable Diffusion · pre-trained SD | attack performance | No clear effect |
| lower filter radius to r_t=1 (ablation) | Naive+F · MS-COCO | AUC: 98.32 -> 90.24 | Hurt |
| raise filter factor s from 0.2 to 0.5 (ablation) | Naive+F · MS-COCO | AUC: 98.32 -> 91.50 | Hurt |
| add high-frequency filter | Naive/SecMI/PIA · fine-tuned SD with defenses (SSei + augmentation) | AUC: +6.14 / +5.10 / +6.86 | Improved |
| adaptive defense suppressing low-frequency components | FreMIA filter · fine-tuned SD | attack effectiveness | Hurt |
Closest papers by meaning
- Towards Black-Box Membership Inference Attack for Diffusion Models (ICML 2025 · similarity 0.93)
- Membership Inference Attacks on Diffusion Models via Quantile Regression (ICML 2024 · similarity 0.93)
- An Efficient Membership Inference Attack for the Diffusion Model by Proximal Initialization (ICLR 2024 · similarity 0.91)
- Membership Inference on Text-to-Image Diffusion Models via Conditional Likelihood Discrepancy (NeurIPS 2024 · similarity 0.90)
- Privacy Attacks on Image AutoRegressive Models (ICML 2025 · similarity 0.89)
- CDI: Copyrighted Data Identification in Diffusion Models (CVPR 2025 · similarity 0.89)
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.
