Transferring Robustness for Graph Neural Network Against Poisoning Attacks

Implementation of paper "Transferring Robustness for Graph Neural Network Against Poisoning Attacks"

by Xianfeng Tang, Yandong Li, Yiwei Sun, Huaxiu Yao, Prasenjit Mitra, Suhang Wang
Published at WSDM 2020, Houston, Texas, USA

Please cite our paper if the model or the paper help:

@inproceedings{tang2020transferring,
	title = {Transferring Robustness for Graph Neural Network Against Poisoning Attacks},
	author={Tang, Xianfeng and Li, Yandong and Sun, Yiwei and Yao, Huaxiu and Mitra, Prasenjit and Wang, Suhang},
	booktitle={ACM Internatioal Conference on Web Search and Data Mining (WSDM)},
	year = {2020}
}

Requirements

Python 3.7 or newer
numpy
tensorflow
scipy

Before running

Please download data.zip and extract all contents to data/.

Run the code

Please run with python main.py.

Contact

Please contact tangxianfeng at outlook.com for any questions.

References

Dataset

Pubmed

We acquire the processed graph from https://github.com/tkipf/gcn/tree/master/gcn/data and put them in data/gcn_data (must be unzip from data.zip to find it). The original datasets can be found here: http://linqs.cs.umd.edu/projects/projects/lbc/.

The original Reddit graph can be found here: http://snap.stanford.edu/graphsage/.

Yelp small & large

We use Yelp Dataset to compile these two datasets.

Code & model design

Meta-learning

The design and implenmentation of meta-learning part is inspired by MAML-TensorFlow and maml.

Graph neural networks

The design of neural networks is inspired by gcn and metattack.

Graph adversarial attacks

We adopt nettack and metattack.

tangxianfeng/PA-GNN