Scholarlink

<p>In this paper, a privacy-preserving federated learning scheme based on multi-key homomorphic encryption is proposed to address the privacy protection challenges faced by load forecasting in distributed smart grids. In view of the collusion attack risks of existing single-key homomorphic encryption schemes and the shortcomings of traditional multi-key schemes in terms of system robustness and communication efficiency, this study innovatively designs the TMK-CKKS multi-key homomorphic encryption scheme, and combines the trusted execution environment with key negotiation technology to build a federated learning framework with strong privacy protection capabilities. This scheme achieves breakthroughs through three key technical innovations: first, the proposed TMK-CKKS scheme effectively prevents collusion attacks between the server and the participants; second, the secure decryption mechanism based on TEE ensures that the system can still decrypt when some nodes are offline, improving the robustness of the system; finally, the optimized decryption algorithm enables the system to only require a single round of communication during the aggregation process, reducing the communication overhead. Through security analysis and comprehensive experimental evaluation, this paper demonstrates that our scheme has high computational efficiency and communication efficiency while ensuring privacy protection.</p>

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

Privacy-preserving federated learning scheme for distributed smart grid based on multi-key homomorphic encryption

  • 期刊论文

  • 发表日期：

    2025年10月7日

查看全文
  • Penglin Zhang,
  • Yong Zhang,
  • Zhaodong Wang,
  • Qiuyao Zhang,
  • Zhenghua Gu

摘要

In this paper, a privacy-preserving federated learning scheme based on multi-key homomorphic encryption is proposed to address the privacy protection challenges faced by load forecasting in distributed smart grids. In view of the collusion attack risks of existing single-key homomorphic encryption schemes and the shortcomings of traditional multi-key schemes in terms of system robustness and communication efficiency, this study innovatively designs the TMK-CKKS multi-key homomorphic encryption scheme, and combines the trusted execution environment with key negotiation technology to build a federated learning framework with strong privacy protection capabilities. This scheme achieves breakthroughs through three key technical innovations: first, the proposed TMK-CKKS scheme effectively prevents collusion attacks between the server and the participants; second, the secure decryption mechanism based on TEE ensures that the system can still decrypt when some nodes are offline, improving the robustness of the system; finally, the optimized decryption algorithm enables the system to only require a single round of communication during the aggregation process, reducing the communication overhead. Through security analysis and comprehensive experimental evaluation, this paper demonstrates that our scheme has high computational efficiency and communication efficiency while ensuring privacy protection.