Federated Learning with Differential Privacy

This project implements a federated learning system with differential privacy using the Flower library. The system is designed to train a neural network model on distributed healthcare data while preserving privacy through differential privacy mechanisms.

Project Overview

The project implements a client-server architecture where:

• A central server coordinates the federated learning process
• Multiple clients (10 in this implementation) train the model on their local data
• Differential privacy is applied to protect sensitive information during the training process

Dataset

The project uses a healthcare dataset containing the following features:

• ABP Diastolic
• ABP Systolic
• Glucose
• Heart Rate
• Respiratory Rate
• Temperature (°F)
• hospital_expire_flag (target variable)

The dataset contains 71,026 rows of patient data.

Project Structure

.
├── data/
│   └── merged_data.csv
├── server.py
├── client.py
├── start_clients.sh
├── requirements.txt
└── README.md

Requirements

• Python 3.7+
• Flower (flwr)
• TensorFlow
• Pandas
• NumPy

Installation

1. Clone the repository:

git clone https://github.com/Gmehta604/Differential-Privacy-Federated-Learning.git

2. Install the required packages:

pip install -r requirements.txt

Usage

1. Start the server:

python server.py

2. Start the clients (in a new terminal):

./start_clients.sh

Implementation Details

Server

• Coordinates the federated learning process
• Aggregates model updates from clients
• Implements differential privacy mechanisms

Clients

• Each client receives a portion of the dataset
• Trains the model locally
• Applies differential privacy to model updates
• Sends encrypted updates to the server

Differential Privacy

The implementation includes differential privacy mechanisms to protect sensitive data during the federated learning process.

Contributing

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

License

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

Acknowledgments

• Flower library for federated learning implementation
• TensorFlow for neural network implementation
• Contributors and maintainers of the project