"Federated Learning for Privacy-Preserving AI in Edge and IoT Devices"

Abstract - This paper presents a study on the application of federated learning (FL) frameworks to enhance privacy-preserving artificial intelligence in edge and Internet of Things (IoT) environments. The core objective is to demonstrate how decentralized machine learning models can be trained collaboratively across distributed devices without sharing raw data, thus addressing major privacy and data security concerns. The research explores key architectural models of federated learning, evaluates optimization algorithms for efficient communication, and integrates secure aggregation techniques to ensure data confidentiality during model updates. Experimental implementations using publicly available healthcare and smart surveillance datasets highlight the performance benefits and trade-offs of FL compared to traditional centralized approaches. Key challenges such as data heterogeneity, communication bottlenecks, and model drift are also analyzed. Results show that federated models maintain competitive accuracy while significantly reducing privacy risks and transmission overhead. The proposed framework has broad applicability in sectors like smart healthcare, agriculture monitoring, and public infrastructure management. This paper concludes that federated learning is a viable and scalable solution for deploying AI across privacy-sensitive, bandwidth-constrained environments, especially within the Indian digital ecosystem. KeyWords: federated learning, edge computing, IoT, privacy-preserving AI, secure aggregation, decentralized intelligence.