Quantum-Safe Hybrid Federated Learning

A robust, multi-layered privacy-preserving framework for Federated Learning (FL). This project integrates Post-Quantum Cryptography (PQC), Homomorphic Encryption (HE), and Differential Privacy (DP) to provide end-to-end security for decentralized machine learning.

Overview

Collaborative training of machine learning models often requires sharing sensitive client data, which introduces significant privacy risks. This framework addresses these vulnerabilities by establishing a "Zero-Trust" environment where:

• Identities are Verified: Using quantum-resistant Dilithium signatures.
• Channels are Secured: Using Kyber-based key encapsulation to protect against quantum-era eavesdropping.
• Aggregation is Blind: Using Paillier Homomorphic Encryption so the server never sees raw weights.
• Users are Anonymized: Using Local Differential Privacy (LDP) to prevent individual data reconstruction.

Key Features

• Standardized PQC: Implementation of NIST-standardized algorithms (ML-KEM/Kyber and ML-DSA/Dilithium).
• Zero-Knowledge Aggregation: Additive homomorphic encryption ensures only aggregated results are ever visible to the server.
• Formal Privacy Guarantees: Noise injection via the Laplace mechanism with configurable privacy budgets (Epsilon).
• Cyber-SOC Dashboard: An interactive React-based dashboard for real-time threat monitoring and system visualization.

Installation

Prerequisites

• Python 3.10+
• Node.js (for the Web GUI)

Setup

1. Clone the repository:

   git clone https://github.com/SamuelJoseph23/quantum-safe-hybrid-ml.git
   cd quantum-safe-hybrid-ml

2. Backend Setup:

   python -m venv venv
   source venv/bin/activate  # On Windows: .\venv\Scripts\activate
   pip install -r requirements.txt

3. Frontend Setup (Optional):

   cd web-gui
   npm install

Quick Start

Core Training (CLI)

To run a headless 10-round federated training session on the Adult Income dataset:

python src/main.py

Interactive Dashboard

To launch the Cyber-SOC visual demo:

1. Start Backend: python src/api.py
2. Start Frontend: cd web-gui && npm run dev
3. Access at http://localhost:5173

Roadmap

• Neural Network Support: Expand from Logistic Regression to deep learning architectures.
• Byzantine Resilience: Implement detection for malicious or poisoned client updates.
• Communication Optimization: Integrate SIMD batching for Homomorphic Encryption to reduce latency.
• TEE Integration: Support for Trusted Execution Environments (Intel SGX) for secure server-side compute.

Contributing

Contributions are welcome! Please see the CONTRIBUTING.md file for setup instructions and code standards.

Security

If you discover a security vulnerability, please refer to our SECURITY.md.

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.