WindVChen/DiffAttack

An unrestricted attack based on diffusion models that can achieve both good transferability and imperceptibility.

37
/ 100
Emerging

Leverages diffusion model latent space to craft semantic perturbations rather than pixel-space manipulations, combining generative synthesis with discriminative attention-disruption techniques. Integrates with Stable Diffusion 2.0 and evaluates across diverse architectures (CNNs, Transformers, MLPs) and defense mechanisms, demonstrating transferability to black-box models while maintaining imperceptibility to human perception.

259 stars.

No Package No Dependents
Maintenance 6 / 25
Adoption 10 / 25
Maturity 9 / 25
Community 12 / 25

How are scores calculated?

Stars

259

Forks

18

Language

Python

License

Apache-2.0

Category

diffusion-adversarial-robustness

Last pushed

Nov 23, 2025

Commits (30d)

0

GitHub
Diffusion Adversarial Robustness · 56 models

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/WindVChen/DiffAttack"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.

Higher-rated alternatives

OPTML-Group/Unlearn-Saliency

[ICLR24 (Spotlight)] "SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in...

49

Shilin-LU/VINE

[ICLR 2025] "Robust Watermarking Using Generative Priors Against Image Editing: From...

41

koninik/DiffusionPen

Official PyTorch Implementation of "DiffusionPen: Towards Controlling the Style of Handwritten...

37

Wuyxin/DISC

(ICML 2023) Discover and Cure: Concept-aware Mitigation of Spurious Correlation

36

bytedance/LatentUnfold

Implementation of paper: Flux Already Knows – Activating Subject-Driven Image Generation without Training

34

Explore Diffusion Models

All categories Trending Diffusion directory Insights