Scholarlink

<p>High-frequency smart meter data streams enable intelligent grid operations but pose significant privacy risks due to their fine-grained nature. This paper proposes gSPF, a three-layer collaborative privacy-preserving framework based on end-edge-cloud architecture. The framework integrates local differential privacy at smart meters, distributed exponential decay sampling at regional edge nodes, and centralized noise fusion for global statistical release. By jointly modeling sampling error and differential privacy noise, gSPF enables coordinated privacy budget allocation across system layers. Experiments on UK-DALE and REFIT datasets demonstrate that gSPF outperforms centralized DP and pure LDP baselines in statistical accuracy, communication efficiency, and downstream task performance, providing a scalable solution for privacy-preserving smart grid analytics.</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

Research on Data Mining and Distributed Feature Protection Algorithms for Fusion Differential Privacy Mechanisms in Intelligent Digital Systems

  • 期刊论文

  • 发表日期：

    2026年4月14日

查看全文
  • Jiaxin Lin,
  • Yangping Chen,
  • Ruiqi Li,
  • Yuetian Huang,
  • Hanye Huang

摘要

High-frequency smart meter data streams enable intelligent grid operations but pose significant privacy risks due to their fine-grained nature. This paper proposes gSPF, a three-layer collaborative privacy-preserving framework based on end-edge-cloud architecture. The framework integrates local differential privacy at smart meters, distributed exponential decay sampling at regional edge nodes, and centralized noise fusion for global statistical release. By jointly modeling sampling error and differential privacy noise, gSPF enables coordinated privacy budget allocation across system layers. Experiments on UK-DALE and REFIT datasets demonstrate that gSPF outperforms centralized DP and pure LDP baselines in statistical accuracy, communication efficiency, and downstream task performance, providing a scalable solution for privacy-preserving smart grid analytics.