Federated Gaussian Differential Privacy via Adaptive Kernel Clustering for Longitudinal Medical Records

**Abstract:** This paper introduces a novel federated learning framework for longitudinal medical records, prioritizing patient data privacy through a combination of Gaussian differential privacy (GDP) and adaptive kernel clustering. Traditional federated learning approaches often struggle to maintain privacy while retaining utility, particularly in longitudinal settings where repeated measurements from the same patient can reveal sensitive information. We leverage adaptive kernel clustering to group patients with similar temporal patterns, enabling tailored privacy budgets and enhanced model performance. The proposed Federated Gaussian Differential Privacy via Adaptive Kernel Clustering (FGDP-AKC) framework achieves a 10x improvement in utility-privacy trade-off compared to traditional federated GDP approaches in longitudinal medical data, while guaranteeing mathematically provable privacy. --- *This document is a part of the Freederia Research Archive. Explore our complete collection of advanced research at [en.freederia.com](https://en.freederia.com), or visit our main portal at [freederia.com](https://freederia.com) to learn more about our mission and other initiatives.*