HondamunigePrasannaSilva/DiffDefence
Official implementation of the paper DiffDefence: defending against adversarial attacks via diffusion models. ICIAP 2023.
This project helps machine learning engineers and researchers protect their image classification models from adversarial attacks without altering the original model. It takes an image that might have been subtly manipulated to trick a classifier and reconstructs it to remove the adversarial perturbation. The output is a "cleaned" image that the original model can classify accurately, even if it was initially vulnerable.
No commits in the last 6 months.
Use this if you need to add a layer of defense to existing image classification models to make them more robust against subtle, malicious alterations of input images.
Not ideal if you are looking for a defense mechanism for non-image data types or if you require extremely fast processing times, as diffusion models can be computationally intensive.
Stars
20
Forks
2
Language
Python
License
—
Category
Last pushed
Feb 01, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/HondamunigePrasannaSilva/DiffDefence"
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...
Shilin-LU/VINE
[ICLR 2025] "Robust Watermarking Using Generative Priors Against Image Editing: From...
WindVChen/DiffAttack
An unrestricted attack based on diffusion models that can achieve both good transferability and...
koninik/DiffusionPen
Official PyTorch Implementation of "DiffusionPen: Towards Controlling the Style of Handwritten...
Wuyxin/DISC
(ICML 2023) Discover and Cure: Concept-aware Mitigation of Spurious Correlation