Homomorphic Encryption Project - Applied Cryptography Group 08

Welcome to the Applied Cryptography - Group 08 project repository. This project demonstrates the implementation of homomorphic encryption for secure healthcare analytics, allowing computations to be performed on encrypted data without decrypting it.

This repository contains three different implementations of homomorphic encryption systems for healthcare analytics, each using different technology stacks and approaches:

Project Versions

1. Microsoft SEAL Version (microsoft-seal/)

• Backend: C++ with Microsoft SEAL library and Crow framework
• Frontend: React with Vite
• Focus: High-performance native implementation with Microsoft's SEAL library
• Use Case: Real-time encrypted data processing and basic homomorphic operations

2. Node-SEAL Version (node-seal/)

• Backend: Node.js with node-seal library and Express
• Frontend: Next.js with TypeScript and modern UI components
• Focus: JavaScript ecosystem implementation with comprehensive UI
• Use Case: Rapid development and modern web-based homomorphic encryption

3. Machine Learning Version (machine-learning/)

• Backend: Python with Concrete ML and TenSEAL libraries
• Models: Logistic Regression and Neural Networks
• Focus: Privacy-preserving machine learning for healthcare predictions
• Use Case: Encrypted diabetes prediction and fully homomorphic ML training

Documentation

Each version has its own detailed documentation:

• Microsoft SEAL Version Documentation - C++ implementation setup and usage
• Node-SEAL Version Documentation - Node.js implementation setup and usage
• Machine Learning Version Documentation - Python ML implementation with FHE

Project Overview

This project is part of the larger initiative "Encrypted Intelligence: Leveraging Homomorphic Encryption for Secure Healthcare Analytics". It demonstrates how to securely process healthcare data in various environments using different homomorphic encryption approaches—enabling computations on encrypted data without exposing sensitive information.

Technology Comparison

Feature	Microsoft SEAL	Node-SEAL	Machine Learning
Language	C++	Node.js/JavaScript	Python
Frontend	React + Vite	Next.js + TypeScript	CLI/Jupyter
Encryption Library	Microsoft SEAL (native)	node-seal (JS bindings)	Concrete ML + TenSEAL
Performance	Highest (native C++)	Good (JavaScript)	Optimized for ML
Use Case	Real-time operations	Web applications	ML predictions

Quick Start

Choose your preferred implementation based on your requirements:

1. For high-performance real-time operations: Use the Microsoft SEAL version

   • Native C++ performance
   • Direct SEAL library integration
   • Ideal for production deployments

2. For modern web development: Use the Node-SEAL version

   • Full JavaScript/TypeScript stack
   • Modern UI with Next.js
   • Rapid prototyping and development

3. For privacy-preserving machine learning: Use the Machine Learning version

   • Python-based ML implementations
   • Diabetes prediction models
   • Fully encrypted training and inference

All implementations provide core homomorphic encryption functionality for secure healthcare analytics, with each optimized for different use cases and deployment scenarios.

Contributors

The following individuals contributed to the development of this project:

• Andrei Visoiu
• Kamil Lipiński
• Luca Nichifor
• Yusuf Serhat Özkan
• Calin Suconicov
• Jakub Mazur
• Tomasz Mizera

License

This project is licensed under the MIT License.