Edge-Enabled Diagnostic Engine with Homomorphic Encryption for RealTime Medical Decision Support

Abstract

The rapid expansion of digital healthcare systems has intensified the demand for secure and privacypreserving data analysis, particularly in sensitive domains such as medical diagnosis. Conventional approaches often process raw patient data directly, increasing exposure to security risks, unauthorized access, and privacy breaches, while also lacking robust authentication mechanisms and failing to ensure data confidentiality during both training and prediction phases. This creates a critical need for integrated frameworks that combine accurate prediction with strong privacy protection and secure system access. To address these challenges, this study proposes a privacy-preserving healthcare prediction system that integrates Machine Learning (ML) with encryption and secure authentication mechanisms. The system is developed using the Flask framework and incorporates Simple Mail Transfer Protocol (SMTP)-based One-Time Password (OTP) verification to ensure secure user login. Data privacy is maintained through Lightweight Privacy-Preserving Machine Learning Encryption (LPME), where dataset features are encrypted using polynomial-based cryptographic operations prior to model training, preventing data leakage. The system supports datasets such as heart disease and hypothyroid conditions, applying preprocessing techniques including normalization, label encoding, and Synthetic Minority Oversampling Technique (SMOTE) to handle class imbalance effectively. For prediction, Gaussian Naive Bayes (GNB) and Extreme Gradient Boosting (XGBoost) models are employed, with XGBoost serving as the primary model due to its ability to capture complex feature interactions and deliver higher predictive performance, while GNB is used for comparative analysis. The system’s performance is evaluated using accuracy, precision, recall, and F1-score.