Complementing the v0.14 release of the OpenDP Library, we’d also like to announce v0.5 of DP Wizard, a user interface that makes it easier to get started with differential privacy. This new version of DP Wizard provides access to the new synthetic data generation functionality of OpenDP v0.14: When running locally you can provide a CSV, specify columns of interest and their bounds, and the software will then infer the relationships between columns and construct a synthetic dataset which shares many characteristics of original, while assuring the privacy of individuals who contributed to the CSV.
The other big new feature since the last blog post is a cloud deployment. Because we really do not want to see anyone’s private data, we do not support file upload here: Instead, you provide a list of column names, and at the end, instead of downloading statistics, you download an unexecuted notebook for your analysis. You can run this notebook in your local environment with private data.
Since the last blog post about DP Wizard there have been a few other improvements:
- Tutorial notes have been expanded, and they can be toggled on and off.
- Light mode / dark mode can also be toggled.
- The code-templating engine has been factored out as a separate repo and package: dp-wizard-templates.
- When selecting the unit of privacy, we’ve added another step to encourage the user to think about the kind of entity being protected.
A full list of changes is in the CHANGELOG.
Try it out, and if you find it useful, let us know your priorities for the next features.