This is the official repository for DPDGC: Differentially Private Decoupled Graph Convolutions for Multigranular Topology Protection (NeurIPS 2023). We study the problem of differential privacy (DP) for GNNs (i.e., DP-GNNs) starting from the definition. Note that a refined camera-ready version will be online soon.

We conducted our code tests on a Linux machine running CentOS 7.9.2009 and Python 3.9.16. For the necessary packages, please refer to requirements.txt. It's important to note that scipy needs to be downgraded to version 1.8.0 to ensure compatibility with autodp due to a recognized issue.

We have consolidated all functionalities within train.py. To execute the code with the default parameters, use the command:

python train.py $dp_method --dataset $dataset --epsilon $epsilon --project $output_folder

Here:

• $dp_method is the desired DP method.
• $dataset is the dataset you wish to test.
• $epsilon is the specified value of epsilon.
• $output_folder is the directory where results will be stored.

Currently, we support the following DP methods

• mlp, mlp-dp
• gap-inf, gap-edp, gap-ndp
• sage-inf, sage-edp, sage-ndp
• dpdgc-inf, dpdgc-edp, dpdgc-ndp, dpdgc-kndp

To reproduce the results shown in our paper and below, you can use the shell scripts in train_scripts with fine-tuned parameters.

Please contact Chao Pan (chaopan2@illinois.edu), Eli Chien (ichien3@illinois.edu) if you have any question.

If you use our code in your work, please cite our paper:

@article{chien2023differentially,
  title={Differentially Private Decoupled Graph Convolutions for Multigranular Topology Protection},
  author={Chien, Eli and Chen, Wei-Ning and Pan, Chao and Li, Pan and {\"O}zg{\"u}r, Ayfer and Milenkovic, Olgica},
  journal={arXiv preprint arXiv:2307.06422},
  year={2023}
}

Please also cite the GAP paper as our code is mainly developed based on their repository.

@inproceedings{sajadmanesh2023gap,
  title={Gap: Differentially private graph neural networks with aggregation perturbation},
  author={Sajadmanesh, Sina and Shamsabadi, Ali Shahin and Bellet, Aur{\'e}lien and Gatica-Perez, Daniel},
  booktitle={USENIX Security 2023-32nd USENIX Security Symposium},
  year={2023}
}

Please also cite the LINKX paper as our DPDGC is developed based on its official implementation.

@article{lim2021large,
  title={Large scale learning on non-homophilous graphs: New benchmarks and strong simple methods},
  author={Lim, Derek and Hohne, Felix and Li, Xiuyu and Huang, Sijia Linda and Gupta, Vaishnavi and Bhalerao, Omkar and Lim, Ser Nam},
  journal={Advances in Neural Information Processing Systems},
  volume={34},
  pages={20887--20902},
  year={2021}
}

Please cite the GPRGNN paper if you use the cSBM synthetic dataset, as our code pertaining to cSBM is directly adopted from their repository.

@inproceedings{
chien2021adaptive,
title={Adaptive Universal Generalized PageRank Graph Neural Network},
author={Eli Chien and Jianhao Peng and Pan Li and Olgica Milenkovic},
booktitle={International Conference on Learning Representations},
year={2021},
url={https://openreview.net/forum?id=n6jl7fLxrP}
}