respark 0.0.4

Generate privacy‑safe synthetic data from production Spark DataFrames

ReSpark

Status: Pre-release (0.0.x)

ReSpark is a Python library built on PySpark for generating privacy-preserving synthetic data from existing Spark DataFrames or schemas. It is designed to run in any environment where Spark is available, whether on a local machine, a cluster, or a cloud platform.

Modern data-driven solutions require realistic datasets for development, testing, and analytics. However, using production data introduces privacy risks and governance challenges. ReSpark provides a privacy-first approach to synthetic data generation, preserving the structure and statistical characteristics of your original data while minimising re-identification risk.

Vision

• Runs Anywhere Spark Runs: Works in any environment where Spark DataFrames are processed, from local setups to large-scale clusters.
• Privacy-First Design: Includes validation reporting to check for residual sensitive information or re-identification risk.
• Relational Integrity: Maintains join consistency with appropriate handling of sensitive and non-sensitive fields.

Installation

pip install respark

This package requires pyspark (Apache-2.0)

Licence

© Crown Copyright 2025 Department for Education
Licensed under the MIT Licence.

Acknowledgements

Built on Apache Spark / PySpark (Apache License 2.0).