Scholarlink

<p>The integration of blockchain and federated learning (FL) has emerged as a promising approach for enabling secure and privacy-preserving analytics in Agricultural Internet of Things (Agri-IoT) systems. However, practical deployment of FL in agricultural environments is challenged by heterogeneous devices, non-IID data, intermittent connectivity, and energy constraints, while centralized machine learning approaches suffer from data leakage risks and single points of failure. This paper presents a Blockchain-Assisted Federated Learning Framework (BA-FLF) for Agri-IoT that combines distributed model training with blockchain-based auditability. Rather than introducing a new consensus protocol, the framework adopts an Agri-IoT-aware hybrid dBFT–PoS consensus adaptation designed to balance scalability, latency, and energy efficiency under resource-constrained conditions. Lightweight on-chain coordination and secure off-chain model storage are employed to reduce system overhead. Simulation-based evaluation under non-IID data settings shows that BA-FLF demonstrates consistent improvements in model accuracy (4–6%), latency (up to 27%), and energy efficiency (up to 22%) compared with baseline federated and centralized learning configurations. These results indicate that BA-FLF offers a promising system-level design for secure and scalable intelligent agriculture applications, while motivating further real-world validation.</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

BA-FLF: blockchain-assisted federated learning framework for secure and privacy-preserving agricultural IoT systems

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • Atul Lal Shrivastava,
  • Rajendra Kumar Dwivedi,
  • Shiva Prakash

摘要

The integration of blockchain and federated learning (FL) has emerged as a promising approach for enabling secure and privacy-preserving analytics in Agricultural Internet of Things (Agri-IoT) systems. However, practical deployment of FL in agricultural environments is challenged by heterogeneous devices, non-IID data, intermittent connectivity, and energy constraints, while centralized machine learning approaches suffer from data leakage risks and single points of failure. This paper presents a Blockchain-Assisted Federated Learning Framework (BA-FLF) for Agri-IoT that combines distributed model training with blockchain-based auditability. Rather than introducing a new consensus protocol, the framework adopts an Agri-IoT-aware hybrid dBFT–PoS consensus adaptation designed to balance scalability, latency, and energy efficiency under resource-constrained conditions. Lightweight on-chain coordination and secure off-chain model storage are employed to reduce system overhead. Simulation-based evaluation under non-IID data settings shows that BA-FLF demonstrates consistent improvements in model accuracy (4–6%), latency (up to 27%), and energy efficiency (up to 22%) compared with baseline federated and centralized learning configurations. These results indicate that BA-FLF offers a promising system-level design for secure and scalable intelligent agriculture applications, while motivating further real-world validation.