PrivGE

This repository contains code for CIKM 2024 paper titled Privacy-Preserving Graph Embedding based on Local Differential Privacy.

Getting Started

Requirements

• pyg 2.2.0
• pytorch 1.12.0
• pybind11 2.10.3

Dataset

Get datasets through the link and put them to the corresponding directories. For example, Cora dataset should be placed into datasets/cora/.

Usage

cd precompute
make

Node Classification

Train with the following command, optional arguments could be found in classification.py.

bash node_class.sh

Link Prediction

Train with the following command, optional arguments could be found in link_pred.py.

bash link_prediction.sh

Citation

Please cite our paper if you use the code in your work:

@inproceedings{10.1145/3627673.3679759,
  author = {Li, Zening and Li, Rong-Hua and Liao, Meihao and Jin, Fusheng and Wang, Guoren},
  title = {Privacy-Preserving Graph Embedding based on Local Differential Privacy},
  year = {2024},
  publisher = {Association for Computing Machinery},
  doi = {10.1145/3627673.3679759},
  booktitle = {Proceedings of the 33rd ACM International Conference on Information and Knowledge Management},
  series = {CIKM '24}
}