International Journal of Scientific Research and Engineering Development

The study presents a comprehensive review of Federated Learning , outlining a practical method for building strong Artificial Intelligence models in the healthcare industry while maintaining patient privacy.A key issue identified is that privacy laws like GDPR and HIPAA often result in medical data being scattered across different organizations, complicating the development of dependable and broadly applicable AI solutions. FL addresses this by allowing teams to train models together without sharing the actual sensitive data.
Research underscores that traditional Federated Learning faces challenges due to non-IID data distributions and susceptibility to privacy inference attacks.This paper examine advanced frameworks that implement a multi-layered defense system apply techniques like Secure Multi-Party Computation, Homomorphic Encryption, and Differential Privacy to tackle these issues. The key advancements discussed affect hierarchical architectures with edge servers to enhance efficiency, dynamic aggregation strategies to manage data heterogeneity, and adaptive privacy budget allocation to achieve an optimal balance between privacy and utility.
The conceptual framework of this study highlights how these enhanced federated learning approaches offer robust privacy safeguards and achieve remarkable diagnostic accuracy, often matching or even exceptional traditional centralized models. However, despite these achievements, the main challenge—especially when using advanced encryption techniques—remains the substantial processing overhead.