Peer Based Mixing for Decentralized Federated Learning

Peer-based decentralized federated learning (DFL) over a stateless mixnet, with full control via FastAPI and a React frontend.

---

Table of Contents

• Features
• Quickstart
• Project Structure
• Useful Commands
• Main Technologies Used

---

Features

• Peer-based DFL simulation using Docker
• Stateless mixnet communication based on
• Federated learning over encrypted transport
• FastAPI backend for node management
• React frontend for UI control
• Inline acknowledgments and message tracking

---

Main Technologies Used

• FastAPI — Python async web framework for backend management
• React.js — Frontend JavaScript library for building user interfaces
• Docker — Containerization for node simulation
• SphinxMix — Mixnet packet format used for anonymous routing
• Scikit-learn — For machine learning models
• Uvicorn — ASGI server used to run FastAPI apps

---

Quickstart

1. Clone the repository

git clone 
cd DFL_PeerBasedMixing

2. Build the Docker image for nodes

docker build -t dfl_node ./node

or for debugging purposes

docker build -t dfl_node ./node --progress=plain --no-cache

3. Install Python requirements

pip install -r requirements.txt

4. Start the FastAPI backend

uvicorn manager.app:app --host 0.0.0.0 --port 8000

Backend runs at http://localhost:8000.

5. Install Node.js and set up the React frontend

sudo apt install npm
npm install --prefix frontend

6. Start the React frontend

npm start --prefix frontend

Frontend runs at http://localhost:3000.

---

Project Structure

DFL_PeerBasedMixing/
├── node/        # Dockerized node logic
├── manager/     # FastAPI backend for node control
├── experiments/ # Jupyter Notebooks and figures
├── metrics/     # Automatically recorded time-series
└── frontend/    # React frontend interface

---