Research map / Adversarial robustness and security / Attacks on text and graphs
Beyond Single-Point Perturbation: A Hierarchical, Manifold-Aware Approach to Diffusion Attacks
Zhijie Wang, Lin Wang, Zhenyu Wen, Cong Wang
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 |
|---|---|---|---|
| replace CFG with prompt-free unconditional self-attention guidance (ablation) | multi-step CFG attack · ImageNet-Compatible, Res-50 | FID: 69.7 -> 62.9 | Improved |
| add self-attention internal guidance (ablation) | HAM · ImageNet-Compatible, Res-50 | transferability | No clear effect |
| add manifold-projected momentum (ablation) | HAM · ImageNet-Compatible, Res-50 | ASR | Improved |
| add dynamic step-wise manifold alignment (ablation) | HAM · ImageNet-Compatible, Res-50 | ASR and FID | Improved |
| hierarchical multi-step attack vs DiffAttack | HAM · ImageNet-Compatible, surrogate Inc-v3 | avg black-box ASR (%): 34.46 -> 56.12 | Improved |
| HAM vs DiffAttack | HAM · ImageNet-Compatible, surrogate Res-50 | avg black-box ASR (%): 60.43 -> 66.63 | Improved |
| HAM vs ACA | HAM · ImageNet-Compatible, surrogate Inc-v3 | avg black-box ASR (%): 62.34 vs 56.12 | Hurt |
| HAM vs DiffAttack | HAM · ImageNet-Compatible, surrogate Inc-v3 | FID: 62.2 -> 64.7 | Hurt |
| HAM vs DiffAttack | HAM · Defended models, Res-50 | avg ASR (%): 60.73 -> 67.11 | Improved |
| HAM vs ACA | HAM · Computational cost | seconds per image: >200 -> 21.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DiffAttack — DiffAttack: Evasion Attacks Against Diffusion-Based Adversarial Purification (NeurIPS 2023) · better on 13 of 14 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
- Adv-Diffusion: Imperceptible Adversarial Face Identity Attack via Latent Diffusion Model (AAAI 2024 · similarity 0.90)
- Variance as a Catalyst: Efficient and Transferable Semantic Erasure Adversarial Attack for Customized Diffusion Models (ICML 2025 · similarity 0.90)
- LoRID: Low-Rank Iterative Diffusion for Adversarial Purification (AAAI 2025 · similarity 0.89)
- Towards Understanding the Robustness of Diffusion-Based Purification: A Stochastic Perspective (ICLR 2025 · similarity 0.89)
- CAT: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models (ICML 2025 · similarity 0.89)
- AdvAD: Exploring Non-Parametric Diffusion for Imperceptible Adversarial Attacks (NeurIPS 2024 · 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.
