/CI-Net

[ICCV 2023] Generative Gradient Inversion via Over-Parameterized Networks in Federated Learning

Primary LanguageJupyter NotebookMIT LicenseMIT

[ICCV 2023] Generative Gradient Inversion via Over-Parameterized Networks in Federated Learning

This is a PyTorch implementation of the paper Generative Gradient Inversion via Over-Parameterized Networks in Federated Learning.

model Architecture

Usage

Install

  • Pytorch + torchvision + piq

Training

  • CIFAR10.ipynb on the CIFAR-10 (resolution: 32 x 32) dataset
  • ImageNet.ipynb on the ImageNet (resolution: 256 x 256) dataset

Visualization

  • For each test, the SSIM value in plotted after the attacking process.
  • The obtained images are also plotted after attacking for visualization.

Results

  • Typical results are as follows. ImageNet Results

Citation

@inproceedings{zhang2023generative,
  title={Generative Gradient Inversion via Over-Parameterized Networks in Federated Learning},
  author={Zhang, Chi and Xiaoman, Zhang and Sotthiwat, Ekanut and Xu, Yanyu and Liu, Ping and Zhen, Liangli and Liu, Yong},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={5126--5135},
  year={2023}
}

License

This project is under the MIT license. See License for details.