Abstract:

Federated Learning (FL) offers a decentralized approach to model training, making it ideal for Internet of Things (IoT) networks that demand both data privacy and timely decision-making. By enabling edge devices to collaboratively learn shared models without transferring raw data, FL strengthens privacy while supporting scalable intelligence across distributed IoT environments. However, traditional centralized learning methods pose significant challenges, including data privacy violations, high latency, and bandwidth inefficiencies due to continuous data transmission to central servers. These limitations hinder real-time decision-making and expose sensitive information to potential breaches. To address these challenges, this paper introduces PRIDE-FL (Privacy-Respecting Intelligent Decision-making using Federated Learning), a novel framework designed to enhance privacy preservation and enable real-time analytics in IoT networks. PRIDE-FL integrates secure aggregation protocols, local model updates, differential privacy, and asynchronous communication to minimize latency and protect user data during model training and inference. PRIDE-FL is deployed across diverse IoT devices such as sensors, smart appliances, and wearables, where edge nodes participate in federated training while ensuring data remains localized. The framework supports intelligent real-time actions based on locally learned insights and periodically updated global models. Experimental results demonstrate that PRIDE-FL achieves improved model accuracy with reduced communication overhead compared to conventional methods. It also maintains strong privacy guarantees, enabling fast and secure decision-making in dynamic IoT settings. This approach highlights the feasibility and advantages of federated learning as a privacy-conscious and efficient solution for modern IoT infrastructures.