Research map / Adversarial robustness and security / Attacks on text and graphs
DADet: Safeguarding Image Conditional Diffusion Models against Adversarial and Backdoor Attacks via Diffusion Anomaly Detection
Hongwei Yu, Xinlong Ding, Jiawei Li, Jinlong Wang, Yudong Zhang, Rongquan Wang, Huimin Ma, Jiansheng Chen
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 reverse step te above 800 (ablation) | DADet · image variation (MS COCO) | precision: recall remains constant while precision decreases | Hurt |
| lower reverse step te below 800 (ablation) | DADet · image variation (MS COCO) | recall: precision unchanged while recall drops | Hurt |
| vary threshold F-hat between 2.5 and 6.5 (ablation) | DADet · image variation (MS COCO) | F1 score: F1 greater than 0.9 across 2.5 to 6.5 | No clear effect |
| adversarial perturbation instead of random noise of equal norm in noise prediction | discrepancy accumulation analysis · reverse process cosine similarity S400 | cosine similarity: 0.91 vs 0.28 | Improved |
| use only homogeneity metric for unconditional models (ablation) | DADet · CIFAR-10 backdoor | precision/recall: performance comparable to Elijah | No clear effect |
| optimize adversarial samples to resemble the condition image (adaptive attack) | DADet · adaptive attack evasion | detection evasion | No clear effect |
| use Stable Diffusion instead of Latent Diffusion for inpainting detection | DADet · Places365 inpainting, Photo guard 8/255 | F1 score: 0.90 -> 0.84 | Hurt |
| use Stable Diffusion instead of Latent Diffusion for inpainting detection | DADet · Places365 inpainting, LDM-R 16/255 | F1 score: 0.97 -> 0.86 | Hurt |
| use DADet with Invisible Backdoor 16/255 instead of 8/255 | DADet · MS COCO backdoor detection, Target Hat | F1 score: 0.98 -> 0.99 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Elijah — Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift (AAAI 2024) · better on 2 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TERD: A Unified Framework for Safeguarding Diffusion Models Against Backdoors (ICML 2024 · similarity 0.90)
- Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift (AAAI 2024 · similarity 0.89)
- UIBDiffusion: Universal Imperceptible Backdoor Attack for Diffusion Models (CVPR 2025 · similarity 0.89)
- UFID: A Unified Framework for Black-box Input-level Backdoor Detection on Diffusion Models (AAAI 2025 · similarity 0.88)
- Step Vulnerability Guided Mean Fluctuation Adversarial Attack against Conditional Diffusion Models (AAAI 2024 · similarity 0.87)
- AdvAD: Exploring Non-Parametric Diffusion for Imperceptible Adversarial Attacks (NeurIPS 2024 · similarity 0.87)
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.
