GNNFingers Reproduction

This repository contains a reproduction of the paper:

You et al. (2024). "GNNFingers: A Fingerprinting Framework for Verifying the Ownership of Graph Neural Networks"

---

Overview

The reproduction includes the following from the paper:

• Train a target GNN model (GCN/GraphSAGE).
• Generate suspect models (fine-tuned, partially retrained, independent).
• Learn fingerprints to query and test models.
• Train a Univerifier classifier to verify ownership.
• Evaluate with Robustness, Uniqueness, Accuracy, ARUC, and ROC-AUC.

---

Packages and pip installs

!pip install torch torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-2.2.0+cu118.html 

---

Results

ARUC: 0.99

ROC-AUC: 1.0

Mean test accuracy: 0.9900990099009901

---

Acknowledgments

Under the supervision of Dr. Yushun Dong and Mr. Bolin Shen

This project utilized text/code generated with the assistance of OpenAI's ChatGPT-3.5 (Nov 30 version, 2022).

---