Privacy-Preserving Sensing in MIMO Networks Using Federated Learning

Introduction: The rapid growth of Internet of Things (IoT) devices and advancements in wireless communication have driven the adoption of multiple-input multiple-output (MIMO) networks for intelligent sensing. However, traditional centralized data processing raises significant privacy concerns, necessitating privacy-preserving alternatives. Method: This study introduces a federated learning (FL)-based distributed sensing architecture for MIMO networks. Each node locally trains a model using its received signal data and transmits only the model updates to a central server. A novel model aggregation strategy has been developed to account for spatial diversity and varying channel conditions in MIMO systems. Results: Simulation results reveal that the proposed FL-MIMO framework achieves sensing accuracy comparable to centralized methods while maintaining raw data privacy. The approach exhibits robustness to non-independent and identically distributed (non-IID) data and asynchronous communication, with negligible performance degradation. Discussion: The findings demonstrate the feasibility of applying federated learning to MIMObased sensing, addressing key challenges such as communication overhead, model convergence, and security against adversarial threats. The method effectively mitigates privacy risks without compromising sensing performance. Conclusion: The proposed FL-MIMO framework provides a practical and secure solution for privacy- preserving sensing in smart environments. By balancing efficiency and privacy, it facilitates scalable and trustworthy deployment of intelligent sensing applications in real-world MIMO networks.