R-GAP
Gradient attack tool
A tool to demonstrate and analyze attacks on private data in machine learning models using gradients
R-GAP: Recursive Gradient Attack on Privacy [Accepted at ICLR 2021]
34 stars
2 watching
2 forks
Language: Python
last commit: almost 2 years ago Related projects:
| Repository | Description | Stars |
|---|---|---|
 superhedgy/attacksurfacemapper | Automates reconnaissance to identify potential attack surfaces of a target network | 1,314 |
 yunqing-me/attackvlm | An adversarial attack framework on large vision-language models | 161 |
 patrickzh/improved-deep-leakage-from-gradients | This project presents an improved method to extract accurate private training data from publicly shared gradients in distributed learning systems. | 144 |
 zhuohangli/ggl | An attack implementation to test and evaluate the effectiveness of federated learning privacy defenses. | 57 |
 zou-group/textgrad | An autograd engine for textual gradients using large language models to backpropagate gradients. | 1,821 |
 ml-postech/gradient-inversion-generative-image-prior | An implementation of a method to invert gradients in federated learning to potentially reveal sensitive client data | 39 |
 ys-zong/foolyourvllms | An attack framework to manipulate the output of large language models and vision-language models | 14 |
 eth-sri/lamp | This is a research project that extracts text from gradients with language model priors using deep learning techniques. | 23 |
 johnkorzhuk/grabient | A tool to generate linear web gradients with a user interface. | 1,990 |
 borealisai/advertorch | A toolbox for researching and evaluating robustness against attacks on machine learning models | 1,308 |
| junyizhu-ai/surrogate_model_extension | A framework for analyzing and exploiting vulnerabilities in federated learning models using surrogate model attacks | 9 |
 lhfowl/robbing_the_fed | This implementation allows an attacker to directly obtain user data from federated learning gradient updates by modifying the shared model architecture. | 23 |
 jind11/textfooler | A tool for generating adversarial examples to attack text classification and inference models | 494 |
| eth-sri/diffai | Trains neural networks to be provably robust against adversarial examples using abstract interpretation techniques. | 218 |
 ai4risk/antifraud | Develops and evaluates machine learning models for detecting financial fraud | 174 |