https://github.com/edisonleeeee/sgattack

Official Repository for "Adversarial Attack on Large Scale Graph." [TKDE 2021]
https://github.com/edisonleeeee/sgattack

adversarial-attacks gcn graph graph-neural-networks sga sgc

Last synced: about 2 months ago
JSON representation

Official Repository for "Adversarial Attack on Large Scale Graph." [TKDE 2021]

• Host: GitHub
• URL: https://github.com/edisonleeeee/sgattack
• Owner: EdisonLeeeee
• License: mit
• Created: 2021-03-16T11:51:18.000Z (almost 4 years ago)
• Default Branch: master
• Last Pushed: 2021-12-21T12:40:14.000Z (about 3 years ago)
• Last Synced: 2023-10-20T19:36:36.369Z (about 1 year ago)
• Topics: adversarial-attacks, gcn, graph, graph-neural-networks, sga, sgc
• Language: Jupyter Notebook
• Homepage:
• Size: 346 KB
• Stars: 9
• Watchers: 2
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# SGA

Official implementation of Simlified Gradient Attack (SGA) in TKDE 2021: [Adversarial attack on large scale graph](https://arxiv.org/abs/2009.03488).

![SGA](./imgs/SGA.png)

# Implementations

+ The PyTorch implementation can be found at [DeepRobust](https://github.com/DSE-MSU/DeepRobust/blob/master/examples/graph/test_sga.py) or [GraphWar](https://github.com/EdisonLeeeee/GraphWar).

# Update

+ 2021/3/20: add tensorflow 2.x support

# Requirements

+ TensorFlow 1.15 (it should work in `1.14`, `1.13`... and `2.x`)

+ Numpy

+ Scipy

+ tqdm

# Datasets

we consider the largest connected component (LCC) of the graph for each dataset, e.g., `data/cora_lcc`.

Cora, Citeseer and Pubmed could be found at https://github.com/danielzuegner/gnn-meta-attack,

reddit could be found at http://snap.stanford.edu/graphsage/

+ cora

+ citeseer

+ pubmed  

+ reddit

# Usage

How to attack a graph neural networks?

+ Train the surrogate model SGC

+ Generate perturbation (edge flips)

+ Attack target models with adversarial examples (SGC, GCN ...)

See `src/demo.ipynb` for a simple example to run our codes.

# Reproduce results in paper

+ Run `src/generate_edges.ipynb` to generate adversarial edges and save them, the adversarial edges can be generated by `SGA`, `GradArgmax`, `Nettack`.

+ Run `src/attack.ipynb` to test the attack performance on `SGC` and `GCN`.

# Cite

If you make use of this code or the SGA algorithm in your work, please cite the following paper:

```bibtex

@article{li2021adversarial,

  title={Adversarial attack on large scale graph},

  author={Li, Jintang and Xie, Tao and Liang, Chen and Xie, Fenfang and He, Xiangnan and Zheng, Zibin},

  journal={IEEE Transactions on Knowledge and Data Engineering},

  year={2021},

  publisher={IEEE}

}

```