Privacy-preserving Collaborative Learning with Automatic Transformation Search

This repository is an implementation of the privacy-preserving algorithm discussed in

Privacy-preserving Collaborative Learning with Automatic Transformation Search

which can be found at https://arxiv.org/abs/2011.12505

Abstract

Collaborative learning has gained great popularity due to its benefit of data privacy protection: participants can jointly train a Deep Learning model without sharing their training sets. However, recent works discovered that an adversary can fully recover the sensitive training samples from the shared gradients. Such reconstruction attacks pose severe threats to collaborative learning. Hence, effective mitigation solutions are urgently desired. In this paper, we propose to leverage data augmentation to defeat reconstruction attacks: by preprocessing sensitive images with carefully-selected transformation policies, it becomes infeasible for the adversary to extract any useful information from the corresponding gradients. We design a novel search method to automatically discover qualified policies. We adopt two new metrics to quantify the impacts of transformations on data privacy and model usability, which can significantly accelerate the search speed. Comprehensive evaluations demonstrate that the policies discovered by our method can defeat existing reconstruction attacks in collaborative learning, with high efficiency and negligible impact on the model performance.

Citation

If you use this code in your research, please cite it.

@inproceedings{gao2021privacy,
  title={Privacy-preserving collaborative learning with automatic transformation search},
  author={Gao, Wei and Guo, Shangwei and Zhang, Tianwei and Qiu, Han and Wen, Yonggang and Liu, Yang},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={114--123},
  year={2021}
}

@article{gao2023automatic,
  title={Automatic Transformation Search Against Deep Leakage from Gradients},
  author={Gao, Wei and Zhang, Xu and Guo, Shangwei and Zhang, Tianwei and Xiang, Tao and Qiu, Han and Wen, Yonggang and Liu, Yang},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2023},
  publisher={IEEE}
}

Code

The search algorithm file that contains can be found at searchalg/. The other important experimental part can be found at benchmark/.

Setup

You can use anaconda to install our setup by running

conda env create -f environment.yml
conda activate ats

Quick Start

Step1 Download Key Files

tiny_data trained network for search

Step2 Automatic Transformation Search

# download a tiny_data trained network for search 
python -u searchalg/batch_generate.py  --arch=ResNet20-4 --data=cifar100 > batch_generate.sh
bash batch_generate.sh
python -u searchalg/search_best.py --arch=ResNet20-4 --data=cifar100

Step3 Privacy-Perserving Evaluation

bash benchmark/run/benchmark_attack.sh

Visualize Result

Input Image	Reconstruction from Inverting Gradients

Transformed Image	Reconstruction from Inverting Gradients

[Model: standard ResNet20, trained on cifar100 data. The image is from the validation set.]

Acknowledgement

We express gratitudes to the great work Inverting Gradients and DLG as we benefit a lot from both their papers and codes.

License

This repository is released under the MIT license.