Privacy-Preserving Federated Learning on Healthcare Data

This project demonstrates how to use Federated Learning (FL) to train a privacy-preserving machine learning model on healthcare data (diabetes prediction), using the Flower framework and PyTorch.

Instead of sending raw patient data to a centralized server, this project enables multiple clients to collaboratively train a model without sharing their data, preserving privacy and data security.

---

What is Federated Learning?

Federated Learning is a machine learning approach where:

• Multiple clients (e.g., hospitals, devices) train a model locally on their private data.
• Only model updates (weights) are shared with a central server.
• The server aggregates these updates to improve a global model.
• Raw data never leaves the clients, ensuring data privacy.

Federated Learning is ideal for sensitive domains like:

• Healthcare
• Finance
• Mobile devices

---

Project Overview

Feature	Details
Task	Binary classification: Predict whether a person has diabetes
Frameworks	PyTorch + Flower
Privacy	Raw healthcare data stays on each client
Model	Feedforward Neural Network (3-layer)
Training	Federated with 2 simulated clients, 10 rounds
Evaluation	Accuracy is aggregated from all clients
Visualization	Global accuracy vs. training rounds plotted after training

---

Dataset

We use the Pima Indians Diabetes Dataset (diabetes.csv), which contains medical diagnostic data for female patients.

• Kaggle Dataset Link
• Features: Glucose, BloodPressure, BMI, Age, etc.
• Target: Outcome (0 = No diabetes, 1 = Diabetes)

---

Project Structure

.
├── model.py               # PyTorch model definition
├── utils.py               # Data loading & client data splitting
├── server.py              # Flower server with accuracy plotting
├── client.py              # Flower client with local training
├── diabetes.csv           # Dataset file
├── federated\_accuracy\_plot.png  # Generated after training
└── README.md              # You are here 

---

How to Run

Make sure you have diabetes.csv in your project folder.

1. Install Dependencies

pip install flwr torch scikit-learn pandas matplotlib

2. Start the Server (Terminal 1)

python server.py

3. Start Client 0 (Terminal 2)

python client.py
# Enter: 0

4. Start Client 1 (Terminal 3)

python client.py
# Enter: 1

Training will run for 10 federated rounds. A graph showing global accuracy vs. round will be displayed after completion and saved as federated_accuracy_plot.png.

---

Output Sample

---

Why Federated Learning?

Traditional ML	Federated Learning
Centralized data storage	Data remains on client devices
Higher risk of data leaks	Improved privacy and compliance
Requires large central dataset	Enables collaboration across institutions

---

References

• Flower Framework
• PyTorch
• Federated Learning Paper (McMahan et al.)

---

Author

Muhammad Asim Hanif