《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (2): 344-352.DOI: 10.11772/j.issn.1001-9081.2023020244

• 人工智能 • 上一篇    

基于生成式对抗网络的联邦学习激励机制

余孙婕1,2, 曾辉1,2, 熊诗雨1,2, 史红周1()   

  1. 1.移动计算与新型终端北京市重点实验室(中国科学院计算技术研究所), 北京 100190
    2.中国科学院大学 计算机科学与技术学院, 北京 100190
  • 收稿日期:2023-03-09 修回日期:2023-04-17 接受日期:2023-04-21 发布日期:2023-08-14 出版日期:2024-02-10
  • 通讯作者: 史红周
  • 作者简介:余孙婕(1998—),女,浙江宁波人,硕士研究生,主要研究方向:信息安全、区块链
    曾辉(1998—),男,江西赣州人,硕士研究生,主要研究方向:信息安全、区块链
    熊诗雨(1999—),女,重庆人,硕士研究生,主要研究方向:信息安全、区块链;
  • 基金资助:
    国家重点研发计划项目(2018YFB1004705)

Incentive mechanism for federated learning based on generative adversarial network

Sunjie YU1,2, Hui ZENG1,2, Shiyu XIONG1,2, Hongzhou SHI1()   

  1. 1.Beijing Key Laboratory of Mobile Computing and Pervasive Device (Institute of Computing Technology,Chinese Academy of Sciences),Beijing 100190,China
    2.College of Computer Science and Technology,University of Chinese Academy of Sciences,Beijing 100190,China
  • Received:2023-03-09 Revised:2023-04-17 Accepted:2023-04-21 Online:2023-08-14 Published:2024-02-10
  • Contact: Hongzhou SHI
  • About author:YU Sunjie, born in 1998, M. S. candidate. Her research interests include information security, blockchain.
    ZENG Hui, born in 1998, M. S. candidate. His research interests include information security, blockchain.
    XIONG Shiyu, born in 1999, M. S. candidate. Her research interests include information security, blockchain.
  • Supported by:
    National Key Research and Development Program(2018YFB1004705)

摘要:

针对当前联邦学习缺乏公平合理的激励机制,难以衡量不同数据量、不同数据质量、不同数据分布的参与节点的联邦学习贡献度等问题,提出一种基于生成式对抗网络(GAN)的联邦学习激励机制。首先,提出融合训练模型的生成式对抗网络实现高精度样本生成;随后,基于融合训练模型的生成式对抗网络实现激励机制的贡献度评估算法,该算法通过联合模型筛选样本并生成数据标签,引入参与节点的本地数据标签分布平衡非独立同分布数据标签对贡献度评估的影响;最后,使用两阶段Stackelberg博弈实现联邦学习激励过程。安全性分析结果表明,所提激励机制在联邦学习过程中保证数据安全和系统稳定。实验结果表明,所提激励机制具备正确性,贡献度评估算法在不同数据量、不同数据质量和不同数据分布的情况下均有较好的性能。

关键词: 联邦学习, 生成式对抗网络, 激励机制, 两阶段Stackelberg博弈, 数据共享

Abstract:

Focused on the current lack of fair and reasonable incentive mechanism for federated learning, and the difficulty in measuring the contribution to federated learning by participant nodes with different data volumes, different data qualities, and different data distributions, a new incentive mechanism for federated learning based on Generative Adversarial Network (GAN) was proposed. Firstly, a GAN with Trained model (GANT) was proposed to achieve high-precision sample generation. Then, the contribution evaluation algorithm of the incentive mechanism was implemented based on GANT. The algorithm filtered samples and generated data labels through the joint model, and introduced the local data labels of the participant nodes to balance the impact of non-independent identically distributed data labels on the contribution evaluation. Finally, a two-stage Stackelberg game was used to realize the federated learning incentive process. The security analysis results show that the proposed incentive mechanism ensures data security and system stability in the process of federated learning. The experimental results show that the proposed incentive mechanism is correct, and the contribution evaluation algorithm has good performance under different data volumes, different data qualities and different data distributions.

Key words: federated learning, Generative Adversarial Network (GAN), incentive mechanism, two-stage Stackelberg game, data sharing

中图分类号: