Scholarlink

This paper presents a blockchain-based incentivised federated learning GAN model developed to enhance privacy, security and efficiency in collaborative training setups. The framework allows multiple participants to collaboratively train a robust model while maintaining privacy. The blockchain ensures that the distribution of rewards is transparent and immutable, thereby increasing the security of the system. The experimental results demonstrate the effectiveness of our approach in protecting privacy, resisting complex attacks, and improving training efficiency through dynamic rewards. It also examines the impact of different reward values on training outcomes and validates the effectiveness of the incentive mechanism. Future research will investigate the potential of advanced cryptography to enhance the security and integrity of blockchain transactions, with the aim of extending the model to multi-task learning, thereby broadening its applicability across domains and larger datasets.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Trusted Federated Learning Scheme for Distributed GAN Model Training

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Jinge Ma,
  • Zixuan Wang,
  • Mingke Chen,
  • Wei Ren

摘要

This paper presents a blockchain-based incentivised federated learning GAN model developed to enhance privacy, security and efficiency in collaborative training setups. The framework allows multiple participants to collaboratively train a robust model while maintaining privacy. The blockchain ensures that the distribution of rewards is transparent and immutable, thereby increasing the security of the system. The experimental results demonstrate the effectiveness of our approach in protecting privacy, resisting complex attacks, and improving training efficiency through dynamic rewards. It also examines the impact of different reward values on training outcomes and validates the effectiveness of the incentive mechanism. Future research will investigate the potential of advanced cryptography to enhance the security and integrity of blockchain transactions, with the aim of extending the model to multi-task learning, thereby broadening its applicability across domains and larger datasets.