Research map / Adversarial robustness and security / Attacks on text and graphs
Universal Adversarial Purification with DDIM Metric Loss for Stable Diffusion
Li Zheng, Liangbin Xie, Jiantao Zhou, He YiMin
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 |
|---|---|---|---|
| raise threshold tau from 4e-3 to 5e-3 (ablation) | UDAP · CelebA-HQ mixed, Anti-DB | ISM: 0.63 -> 0.28 | Hurt |
| lower tau from 4e-3 to 2e-3 (ablation) | UDAP · CelebA-HQ mixed, Anti-DB | time(s): 3 -> 18 | Hurt |
| lower tau from 4e-3 to 3e-3 (ablation) | UDAP · CelebA-HQ mixed, Anti-DB | ISM: 0.63 -> 0.66 | No clear effect |
| use dynamic epochs (tau=4e-3) instead of fixed 100 epochs (ablation) | UDAP · CelebA-HQ mixed, Anti-DB | time(s): 15 -> 3 | Improved |
| use dynamic epochs instead of fixed 50 epochs (ablation) | UDAP · CelebA-HQ mixed, Anti-DB | ISM: 0.22 -> 0.63 | Improved |
| apply UDAP to clean images (ablation) | UDAP · CelebA-HQ Anti-DB | ISM: 0.63 -> 0.62 | No clear effect |
| apply UDAP to adversarial images (ablation) | UDAP · CelebA-HQ Anti-DB | ISM: 0.40 -> 0.62 | Improved |
| purify with mismatched SD v1.4 for v2.1 attack (ablation) | UDAP · CelebA-HQ Anti-DB | ISM: 0.62 -> 0.61 | No clear effect |
| purify with mismatched SD v2.1 for v1.4 attack (ablation) | UDAP · CelebA-HQ Anti-DB | ISM: 0.64 -> 0.63 | No clear effect |
| apply UDAP with prompt p2 (ablation) | UDAP · CelebA-HQ Anti-DB | ISM: 0.13 -> 0.44 | Improved |
| UDAP vs DiffPure under MetaCloak | UDAP · CelebA-HQ | ISM: 0.31 -> 0.53 | Improved |
| UDAP vs GridPure under Anti-DF | UDAP · CelebA-HQ | ISM: 0.46 -> 0.51 | Improved |
Closest papers by meaning
- LoRID: Low-Rank Iterative Diffusion for Adversarial Purification (AAAI 2025 · similarity 0.91)
- Towards Understanding the Robustness of Diffusion-Based Purification: A Stochastic Perspective (ICLR 2025 · similarity 0.91)
- Divide and Conquer: Heterogeneous Noise Integration for Diffusion-based Adversarial Purification (CVPR 2025 · similarity 0.90)
- Diffusion-based Adversarial Purification from the Perspective of the Frequency Domain (ICML 2025 · similarity 0.90)
- ADBM: Adversarial Diffusion Bridge Model for Reliable Adversarial Purification (ICLR 2025 · similarity 0.89)
- DiffBreak: Is Diffusion-Based Purification Robust? (NeurIPS 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.
