Federated Learning-Based Privacy-Preserving Electronic Healthcare Data Management Using Fuzzy Logic

Abstract

The widespread adoption of electronic health records (EHRs) has significantly advanced the healthcare landscape by promoting extensive data exchange and fostering collaboration across institutions. Nonetheless, this progress has heightened concerns regarding the privacy of sensitive patient data. Federated Learning (FL) has gained traction as a decentralized approach that allows for collaborative model training across distributed systems without sharing raw patient information. This paper introduces an innovative FL framework that incorporates a Multilayer Perceptron (MLP) as its central predictive model, specifically designed for healthcare-related tasks. Although MLP’s offer strong capabilities for modelling intricate health data, FL continues to face issues such as variations in data distribution and ambiguity in assessing the value of individual client updates. To mitigate these challenges, the proposed approach employs fuzzy logic-based trust evaluation, which quantifies the reliability of client contributions by analysing data integrity and behavioural patterns. This mechanism supports more resilient and secure model aggregation by filtering out unreliable or potentially harmful updates. Experimental results using benchmark healthcare datasets show that our MLP-driven FL framework enhances predictive performance, reduces communication load and upholds high standards of data privacy.