Scholarlink

<p>The explosive growth of the Internet of Things (IoT) has introduced significant challenges in data privacy, communication overhead, and real-time analytics. Traditional centralised approaches to data processing not only raise privacy concerns but also introduce high latency and scalability limitations. This paper proposes a novel Hierarchical Federated Edge Learning (HFEL) framework that seamlessly integrates Edge Computing and Federated Learning (FL) to enable privacy-preserving, low-latency, and scalable IoT analytics. Unlike conventional flat FL approaches, the HFEL architecture adopts a multi-tiered aggregation model that supports cluster-level and global-level learning while incorporating ensemble learning at the edge. It leverages adaptive learning rates and a lightweight differential privacy mechanism to enhance model robustness and ensure privacy guarantees without compromising accuracy. The proposed system is evaluated on real-world IoT datasets in three key domains: smart cities, industrial automation, and healthcare. Experimental results show that HFEL achieves up to 31.8% improvement in model accuracy, reduces latency by 27.5%, and lowers communication overhead by 36.2% compared to baseline FL and centralised architectures. These improvements were consistently validated against established baselines, including FedAvg, FedProx, and Scaffold, ensuring that claims are benchmarked against state-of-the-art methods. The framework also includes a fault tolerance mechanism for handling intermittent connectivity and missing updates, which are common in real-world edge environments. Furthermore, it demonstrates resilience under non-IID data distributions and heterogeneous edge device configurations. This work not only provides theoretical modelling and algorithmic design but also offers practical insights into deployment strategies. By integrating federated learning with edge computing hierarchically and adaptively, HFEL provides a scalable, efficient, and secure foundation for future IoT 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

Edge computing and federated learning for privacy-preserving IoT analytics

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月5日

查看全文
  • Mohammed Naif Alatawi

摘要

The explosive growth of the Internet of Things (IoT) has introduced significant challenges in data privacy, communication overhead, and real-time analytics. Traditional centralised approaches to data processing not only raise privacy concerns but also introduce high latency and scalability limitations. This paper proposes a novel Hierarchical Federated Edge Learning (HFEL) framework that seamlessly integrates Edge Computing and Federated Learning (FL) to enable privacy-preserving, low-latency, and scalable IoT analytics. Unlike conventional flat FL approaches, the HFEL architecture adopts a multi-tiered aggregation model that supports cluster-level and global-level learning while incorporating ensemble learning at the edge. It leverages adaptive learning rates and a lightweight differential privacy mechanism to enhance model robustness and ensure privacy guarantees without compromising accuracy. The proposed system is evaluated on real-world IoT datasets in three key domains: smart cities, industrial automation, and healthcare. Experimental results show that HFEL achieves up to 31.8% improvement in model accuracy, reduces latency by 27.5%, and lowers communication overhead by 36.2% compared to baseline FL and centralised architectures. These improvements were consistently validated against established baselines, including FedAvg, FedProx, and Scaffold, ensuring that claims are benchmarked against state-of-the-art methods. The framework also includes a fault tolerance mechanism for handling intermittent connectivity and missing updates, which are common in real-world edge environments. Furthermore, it demonstrates resilience under non-IID data distributions and heterogeneous edge device configurations. This work not only provides theoretical modelling and algorithmic design but also offers practical insights into deployment strategies. By integrating federated learning with edge computing hierarchically and adaptively, HFEL provides a scalable, efficient, and secure foundation for future IoT analytics.