Research map / Adversarial robustness and security / Attacks on text and graphs
NatADiff: Adversarial Boundary Guidance for Natural Adversarial Diffusion
Max Collins, Jordan Vice, Tim French, Ajmal Saeed Mian
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 |
|---|---|---|---|
| use NatADiff instead of constrained attacks PGD/AA | RN-50 surrogate · ImageNet avg transfer ASR | ASR (%): 17.6 / 18.4 -> 68.2 | Improved |
| use NatADiff (untargeted) instead of AdvClass (untargeted) | RN-50 surrogate · ImageNet avg ASR | ASR (%): 45.7 -> 68.2 | Improved |
| use untargeted similarity targeting vs random targets | NatADiff RN-50 surrogate · ImageNet avg ASR | ASR (%): 56.8 -> 68.2 | Improved |
| targeted vs untargeted NatADiff on ViT-H surrogate | NatADiff · ImageNet avg ASR | ASR (%): untargeted 69.7 vs targeted 73.2 (targeted IS 15.3 vs 31.9) | Mixed |
| add classifier augmentations (ablation) | NatADiff RN-50 surrogate, targeted · ImageNet avg ASR | ASR (%): 48.7 -> 56.8 | Improved |
| add classifier augmentations (ablation) | NatADiff RN-50 surrogate, targeted · ImageNet image quality | FID-Val: 56.7 -> 66.5 | Hurt |
| add classifier augmentations (ablation) | NatADiff RN-50 surrogate, targeted · ImageNet-A alignment | FID-A: 81.7 -> 77.3 | Improved |
| increase boundary guidance mu from 0.0 to 0.5 (ablation) | NatADiff RN-50 surrogate · ImageNet avg ASR | ASR (%): 52.1 -> 69.5 | Improved |
| increase boundary guidance mu (ablation) | NatADiff RN-50 surrogate · FID-A | FID-A: lowest 77.3 at mu=0.2; 80.1 at mu=0.5 | Mixed |
| change intersection prompt template (ablation) | NatADiff · avg ASR across 8 templates | ASR (%): varies by less than 7.7% | No clear effect |
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.
- DiffAttack — DiffAttack: Evasion Attacks Against Diffusion-Based Adversarial Purification (NeurIPS 2023) · better on 1 of 2 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
- Intriguing Properties of Diffusion Models: An Empirical Study of the Natural Attack Capability in Text-to-Image Generative Models (CVPR 2024 · similarity 0.88)
- DiffAdvMAP: Flexible Diffusion-Based Framework for Generating Natural Unrestricted Adversarial Examples (ICML 2025 · similarity 0.88)
- Structure-Guided Adversarial Training of Diffusion Models (CVPR 2024 · similarity 0.86)
- Towards Understanding the Robustness of Diffusion-Based Purification: A Stochastic Perspective (ICLR 2025 · similarity 0.86)
- Sample-specific Noise Injection for Diffusion-based Adversarial Purification (ICML 2025 · similarity 0.85)
- ADBM: Adversarial Diffusion Bridge Model for Reliable Adversarial Purification (ICLR 2025 · similarity 0.85)
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.
