A Federated Learning Framework for Energy-Aware and Interference-Conscious Cognitive Radio Spectrum Management

Abstract

This research introduces a federated learning (FL) scheme for energy-efficient and interference-aware spectrum management in Cognitive Radio Networks (CRNs). The proposed framework utilizes decentralized machine learning, allowing edge secondary users’ devices to collaboratively train spectrum management models while preserving raw data privacy, alleviating privacy concerns and lowering communication overhead. The framework applies energy-efficient scheduling and interference management to optimize resource utilization while minimizing the consumption of energy and spectrum interference. Experimental results validate the framework’s capabilities to reduce energy usage and detection accuracy while maintaining effective spectrum utilization even in hostile, noisy environments. Furthermore, adaptive model compression and communication strategies enhance the latency and bandwidth requirement, resulting in a 64% reduction in data transmission costs. The system is fair across varying network sizes while demonstrating robust scalability with minimal accuracy loss in larger networks. The proposed approach shows promise for CRNs due to their dynamic and heterogeneous nature, thus enhances the case for integrating 6G and IoT-based infrastructure into future wireless networks