Enhancing the privacy of eHealth platforms using Federated Learning

Abstract

The rapid digital transformation in the healthcare sector has led to the widespread adoption of eHealth platforms, which enable the remote collection, processing, and analysis of sensitive medical data. Although these platforms offer significant benefits in terms of efficiency and accessibility, they also raise critical concerns about data privacy and security. Federated Learning (FL) has emerged as a promising decentralized paradigm for collaborative machine learning, allowing multiple clients to train a shared global model without directly sharing their raw data. However, despite its distributed nature, FL remains vulnerable to various privacy threats, such as inference attacks and data reconstruction. In this report, we propose a privacy-preserving FL approach that integrates three complementary techniques: FedProx, Advanced Encryption Standard (AES), and Secure Aggregation (SA). The approach investigates the integration of transfer learning and fine-tuning using the Brain Tumor MRI Dataset. The results obtained showed that the proposed approach can significantly reduce privacy risks while delivering competitive predictive accuracy which is a compromise between accuracy and privacy preservation.