Communication-Efficient Federated Learning in Industrial IoT — A Framework for Real-Time Threat Detection and Secure Device Coordination
Year: 2025
Type: article
Abstract: Industrial Internet of Things (IIoT) technology development speed created a necessity for machine learning frameworks which provide secure operations along with efficient communication and ready scalability. The existing centralized approach proves inappropriate for IIoT because it faces limitations from bandwidth limitations and privacy issues in addition to cyber risks. We develop Communication-Efficient Federated Learning (CEFL) framework specifically designed for IIoT operations because it provides real-time intelligence capabilities with lower communication costs and improved security measures. CEFL implements an automated operation where edge devices use local datasets to conduct training tasks in each cycle. The limited bandwidth necessitates devices to use gradient sparsification and quantization techniques which reduces the size of update transmissions. Dependable user updates get collected securely on the central server through differential privacy techniques which protect sensitive information. The system implements an adjusting scheduling framework that adjusts device contribution equilibrium with energy capacity as well as network conditions and trust ratings therefore maximizing resource deployment and providing continuous performance despite hardware outages. The system includes a threat detection module which tracks gradient variations to detect and trigger the removal of potential harmful devices immediately. The system pipeline that includes local optimization and efficient gradient handling together with secure aggregation and adaptive scheduling and proactive threat detection has been mathematically proven for its robust and efficient operation. The experimental tests conducted within simulated IIoT network environments demonstrate that the developed framework reduces communication expenses while maintaining both the model accuracy and security performance. The design of CEFL recognizes and overcomes main IIoT obstacles by delivering adaptable lightweight solutions which work well in complicated industrial conditions. Trust-based device coordination along with proactive anomaly detection leads to an autonomous and resilient network structure which prepares industrial intelligence systems for operation reliability improvement. The proposed framework creates a solid basis for extending digital industrial intelligence that involves energy-efficient federated learning as well as blockchain-based trust systems and multi-domain IIoT operations which drive next-generation industrial intelligence.
Authors: Nutan Gusain, Himanshu Sharma
Institutions: Galgotias University, Sharda University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4410417355