Skip to content

Latest commit

 

History

History
13 lines (8 loc) · 568 Bytes

README.md

File metadata and controls

13 lines (8 loc) · 568 Bytes

Deep Leakage From Gradients

This repository includes an unofficial implementation of the paper entitled: "Deep Leakage from Gradients"

How to use the code

To run the DLG algorithm on one of the 50,000 images in the CIFAR100 dataset use:

  python3 main.py --index <image index ranging from 0 to 49,999>

To visualize the gradient matching loss when adding gaussian noise to the gradients use:

  python3 noise.py -- index <image index ranging from 0 to 49,999  

All the dependencies needed to run the code can be found at requirements.txt