Federated Deep Learning for Privacy-Preserving Real-Time Decision Making in Distributed AI Systems

Abstract

In the era of ubiquitous data generation and stringent privacy regulations, Federated Learning (FL) has emerged as a transformative approach for enabling collaborative model training without centralized data collection. This paper presents a novel framework for privacy-preserving AI in real-time decision-making scenarios within distributed deep learning environments. The proposed architecture leverages edge computing to process data locally on user devices, thereby preserving data confidentiality and minimizing communication overhead. By integrating secure model aggregation mechanisms, the system ensures both performance and privacy are maintained even under adversarial conditions. Experimental results based on simulated deployments demonstrate that the federated approach achieves near-centralized accuracy with significantly reduced privacy risks and latency, validating its applicability for critical applications in healthcare, smart infrastructure, and industrial automation.