Scholarlink

<p>Wireless body area networks (WBANs) in healthcare face profound challenges in achieving secure, real-time data aggregation while preserving strict privacy under severe resource constraints, such as limited computational power, energy budgets, and bandwidth, where centralized systems expose sensitive physiological data to breaches and single points of failure. FL-BlockNet addresses this problem through a novel hybrid framework integrating federated learning (FL) and blockchain technology, with contributions including: (1) priority-based aggregation that assigns higher weights to emergency data for rapid critical alert processing; (2) tailored differential privacy (DP) with type-specific noise parameters; and (3) a lightweight permissioned blockchain using Practical Byzantine Fault Tolerance (PBFT) consensus and smart contracts for fine-grained access control and tamper-proof logging of AES-256-encrypted model updates and SHA-256 hashes. Rigorous evaluation on eICU-CRD, MHEALTH, and Synthetic WBAN datasets reveals FL-BlockNet superior findings: 45.443 ms latency, 4.72% faster than baselines; F1-scores up to 0.952; 41% lower computational overhead; and minimal security overhead (4.69% energy impact). Energy consumption (962.71 µJ) exceeds the target by 113%, but optimizations like 40% model pruning and 8-bit quantization reduce it to &#xa0;430 µJ. Scalability confines latency to 170.7 ms at N=100, but FL-BlockNet rigorous equilibrium of real-time efficacy, privacy fortification, and emergency acuity renders it a seminal solution for small-to-medium WBANs in healthcare, with prospects for broader deployment through advanced parallel transmission and adaptive protocols</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

Fl-blocknet enables privacy-preserving data aggregation in wireless body area networks

  • 期刊论文

  • 发表日期：

    2025年12月12日

查看全文
  • Ben Othman Soufiane

摘要

Wireless body area networks (WBANs) in healthcare face profound challenges in achieving secure, real-time data aggregation while preserving strict privacy under severe resource constraints, such as limited computational power, energy budgets, and bandwidth, where centralized systems expose sensitive physiological data to breaches and single points of failure. FL-BlockNet addresses this problem through a novel hybrid framework integrating federated learning (FL) and blockchain technology, with contributions including: (1) priority-based aggregation that assigns higher weights to emergency data for rapid critical alert processing; (2) tailored differential privacy (DP) with type-specific noise parameters; and (3) a lightweight permissioned blockchain using Practical Byzantine Fault Tolerance (PBFT) consensus and smart contracts for fine-grained access control and tamper-proof logging of AES-256-encrypted model updates and SHA-256 hashes. Rigorous evaluation on eICU-CRD, MHEALTH, and Synthetic WBAN datasets reveals FL-BlockNet superior findings: 45.443 ms latency, 4.72% faster than baselines; F1-scores up to 0.952; 41% lower computational overhead; and minimal security overhead (4.69% energy impact). Energy consumption (962.71 µJ) exceeds the target by 113%, but optimizations like 40% model pruning and 8-bit quantization reduce it to  430 µJ. Scalability confines latency to 170.7 ms at N=100, but FL-BlockNet rigorous equilibrium of real-time efficacy, privacy fortification, and emergency acuity renders it a seminal solution for small-to-medium WBANs in healthcare, with prospects for broader deployment through advanced parallel transmission and adaptive protocols