LDP Toolbox: Exploring Utility and Attackability Tradeoffs in Local Differential Privacy

LDP Toolbox is a Python package for analyzing, comparing, and visualizing Local Differential Privacy (LDP) protocols and their trade-offs between utility, privacy, and attackability.

This toolbox provides:

• 📊 Interactive dashboards powered by Dash
• ⚙️ Protocol implementations for frequency estimation tasks
• 🗂️ Visual tools to compare utility loss (e.g., MSE, KL-divergence), attackability, and privacy budget ε
• 📈 Upload your own data to explore privacy-utility trade-offs

---

🚀 Installation

LDP Toolbox is available on PyPI. Install it with:

pip install ldp-toolbox

⚡ Usage

After installation, you can launch the dashboard in two ways:

✅ Option 1 — Using the CLI (recommended)

Run directly from the terminal:

ldp-toolbox

✅ Option 2 — Using Python module

Alternatively, you can run it as a module:

python -m ldp_toolbox.toolbox.app

Or if you prefer, you can embed the app in your own code:

from ldp_toolbox.toolbox.app import app

if __name__ == "__main__":
    app.run(debug=True)

📁 Project Structure

• ldp_toolbox/
  • protocols/ — Core LDP protocol implementations
  • toolbox/ — Dash front-end app (assets/, pages/, app.py)

Example datasets (data/) are provided in this repository for demonstration and local testing, but are not shipped with the PyPI package.

🎥 Demonstration video

A recorded demonstration video is available at: Demo

🤝 Contributing

LDP-Toolbox is a work in progress, and we expect to release new versions frequently, incorporating feedback and code contributions from the community.

1. Fork this repo.
2. Create a feature branch.
3. Submit a pull request.

---

📬 Contact Authors:

• Haoying Zhang: haoying.zhang [at] inria [dot] fr
• Abhishek K. Mishra: abhishek.mishra [at] inria [dot] fr
• Héber H. Arcolezi: heber.hwang-arcolezi [at] inria [dot] fr

📝 License

This project is licensed under the MIT License.