Federated Learning Implementation to enhance real-time efficiency, and privacy decision-making in Internet of Things Network

Federated Learning (FL) enables distributed model training across Internet of Things (IoT) devices while preserving data privacy. However, traditional FL methods face challenges in real-time efficiency and privacy risks due to direct model updates and potential inference attacks. Existing FL approaches often suffer from communication overhead, model degradation, and privacy vulnerabilities, limiting their effectiveness in dynamic IoT environments. These challenges hinder secure decision-making and degrade overall system performance. Differential Privacy (DP) ensures data confidentiality by adding noise to local updates, preventing data reconstruction, while Secure Aggregation (SA) enables encrypted model updates, safeguarding information from adversarial access. This dual-layer protection enhances privacy without compromising learning efficiency. The proposed DP-SA method is integrated into FL for real-time IoT applications, improving security and decision-making without exposing sensitive data. This approach ensures robust model training while minimizing computational and communication overhead, making it ideal for resource-constrained IoT networks. Experimental results demonstrate that the DP-SA framework significantly improves privacy preservation, reduces latency, and enhances model accuracy in real-time IoT decision-making. The proposed method outperforms existing FL approaches, providing a more secure and efficient solution for IoT networks.