LDP-SynthData

Local Differential Privacy Synthetic-Data Generator

A single-file CLI that generates privacy-preserving synthetic CSVs via local differential privacy (Laplace & randomized response).

What it is


A CLI that takes your sensitive tabular CSV and produces a privacy-preserving synthetic twin via local-DP (e.g. randomized response + local noise).

Why you need it

• Privacy regulations (GDPR, CCPA) are forcing every team to explore synthetic data.
• Local DP (each user perturbs on-device) is hotter research than “central DP.”
• Nobody’s distilled it to a one-file script—existing toolchains are multi-repo or Jupyter mashups.

Key features

python dpgen.py
--input data.csv
--epsilon 1.0
--mechanism laplace
--output synthetic.csv

• Supports categorical & numeric columns
• Pluggable noise mechanisms (Laplace, Gaussian, randomized-response)
• Minimal deps: numpy, pandas, scikit-learn

Usage example

python LDP-SynthData.py
--input your_data.csv
--epsilon 1.0
--mechanism all
--output synthetic.csv

Dependencies

pip install pandas numpy