The rapid proliferation of the Industrial Internet of Things (IIoT) has enabled intelligent automation, real-time analytics, and predictive maintenance across modern industries. However, the massive volume of data generated by distributed IIoT devices poses critical challenges for data privacy, communication overhead, and computational efficiency. Traditional centralized learning approaches are often unsuitable for such scenarios due to data sensitivity and bandwidth limitations. To address these issues, this paper proposes a federated edge learning (FL)-based framework for IIoT applications that enables collaborative model training across distributed edge devices without sharing raw data. The proposed architecture leverages edge and fog computing layers to coordinate local model aggregation, reduce latency, and optimize communication cost. In addition, a privacy-preserving aggregation mechanism is incorporated to ensure a secure model update against potential inference attacks. Experimental evaluation of IIoT benchmark datasets demonstrates that the proposed method achieves accuracy comparable to centralized learning while reducing communication overhead by up to 35% and preserving data confidentiality. The results validate the potential of FL as a scalable and privacy-aware solution for next-generation industrial intelligence, paving the way for secure and efficient IIoT deployments in 5G and beyond networks. The source code is available at https://github.com/minh124201/FedFault.git.

Federated Learning,Industrial Internet of Things,edge computing,Privacy Preservation,Distributed Machine Learning,5G/6G