README.md

metadata

title: Privacy Preserving Machine Learning
emoji: 🏆
colorFrom: gray
colorTo: blue
sdk: gradio
sdk_version: 6.9.0
app_file: app.py
pinned: false
license: mit
short_description: Project demonstrates privacy-preserving techniques for ML

Privacy-Preserving Machine Learning Demo

🔒 Project Overview

This project demonstrates privacy-preserving techniques for machine learning on sensitive healthcare data. It implements:

• SHA-256 Hashing for direct identifiers (SSN)
• Pseudonymization for names
• K-Anonymity Generalization for DOB and income
• Laplace Noise (Differential Privacy) for numerical values
• Differentially Private ML Training using IBM's diffprivlib

📁 Files Included

File	Description
app.py	Gradio web interface (main entry point for HF Spaces)
privacy_ml_solution.py	Core ML pipeline with all privacy techniques
requirements.txt	Python dependencies
Assignment2Dataset-1_encrypted.csv	The encrypted/anonymized dataset
model_comparison_results.csv	Performance metrics comparing models
Privacy_Preserving_ML_Report.docx	Comprehensive academic report
Technical_Documentation.docx	Code and library documentation

---

🚀 Deploying to Hugging Face Spaces (Step-by-Step for Beginners)

Step 1: Create a Hugging Face Account

1. Go to huggingface.co
2. Click "Sign Up" in the top right
3. Fill in your details and verify your email

Step 2: Create a New Space

1. Once logged in, click your profile picture → "New Space"
2. Fill in the form:
   • Owner: Select your username
   • Space name: e.g., privacy-ml-demo
   • License: MIT
   • SDK: Select Gradio
   • Hardware: Keep as "CPU Basic" (free)
3. Click "Create Space"

Step 3: Upload Your Files

Option A: Using the Web Interface (Easiest)

1. In your new Space, click the "Files" tab
2. Click "Add file" → "Upload files"
3. Upload these files:
   • app.py (REQUIRED - this is the entry point)
   • requirements.txt (REQUIRED)
   • Assignment2Dataset-1_encrypted.csv (optional sample data)
4. Wait for the build to complete (~2-3 minutes)

Option B: Using Git (For More Control)

# Clone your space
git clone https://huggingface.co/spaces/YOUR_USERNAME/privacy-ml-demo
cd privacy-ml-demo

# Copy your files into the directory
cp /path/to/app.py .
cp /path/to/requirements.txt .

# Commit and push
git add .
git commit -m "Initial upload"
git push

Step 4: Wait for Build

1. Click the "App" tab to see your space building
2. Watch the logs for any errors
3. Once complete, your app will be live!

Step 5: Test Your App

1. Your app is now live at: https://huggingface.co/spaces/YOUR_USERNAME/privacy-ml-demo
2. Upload a CSV file and adjust the epsilon slider
3. Click "Run Privacy Analysis" to see results

---

⚙️ Local Testing (Before Deployment)

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run the app
python app.py

# Open http://localhost:7860 in your browser

---

🔧 Troubleshooting

"Build failed" Error

• Check that requirements.txt has correct package names
• View build logs for specific error messages

"App crashed" Error

• Ensure app.py has demo.launch() at the bottom
• Check for syntax errors in your code

Slow Loading

• Free tier spaces "sleep" after inactivity
• First load takes ~30 seconds to wake up

Memory Issues

• Reduce n_estimators in RandomForest
• Use smaller test datasets

---

📊 Understanding the Results

Metric	What it Means
Accuracy	% of correct predictions
F1 Score	Balance of precision and recall
Epsilon (ε)	Privacy budget - lower = more privacy

Privacy Level Guide

• ε = 0.1-0.5: Very high privacy, some accuracy loss
• ε = 1.0: Balanced (recommended)
• ε = 5.0+: Lower privacy, minimal accuracy impact

---

📚 Learn More

• Differential Privacy Explained
• IBM diffprivlib Documentation
• Gradio Documentation
• Hugging Face Spaces Guide

---

📝 License

MIT License - Feel free to use and modify for your projects.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference