Personalized Federated Graph Learning (PFL_Graph)

• This repository presents a unified framework for personalized federated learning on graph-structured data, mainly for node-level tasks.
• Part of these codes are motivated by PFL-Non-IID
• To run our ConFGL implementation, please select "grace" model and "Ditto" FL algorithm in run.sh script.
• place your customized dataset (e.g., network graph for intrusion detection) in dataset directory.

Citation

• please cite our work if you appreciate it, thank you a lot!!!

@INPROCEEDINGS{10448483,
  author={Mao, Qinghua and Lin, Xi and Su, Xiu and Li, Gaolei and Chen, Lixing and Li, Jianhua},
  booktitle={ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, 
  title={Prompting Label Efficiency in Federated Graph Learning Via Personalized Semi-Supervision}, 
  year={2024},
  pages={6605-6609},
  keywords={Training;Degradation;Federated learning;Graphics processing units;Self-supervised learning;Signal processing;Throughput;Personalized Federated Learning;Graph Neural Networks;Label Deficiency;Contrastive Learning},
  doi={10.1109/ICASSP48485.2024.10448483}}

@article{mao2025fecograph,
  title={FeCoGraph: Label-Aware Federated Graph Contrastive Learning for Few-Shot Network Intrusion Detection},
  author={Mao, Qinghua and Lin, Xi and Xu, Wenchao and Qi, Yuxin and Su, Xiu and Li, Gaolei and Li, Jianhua},
  journal={IEEE Transactions on Information Forensics and Security},
  year={2025},
  publisher={IEEE}
}