Release Date: July 1, 2025
Verify differential privacy using SDVerified. Our new, SDVerified feature allows you to measure the differential privacy of your synthesizer, and record the fact that you’ve verified it. Use this for any differential privacy synthesizer (although you can run the evaluation on its own for any other synthesizer too).
Extract timezones from your data. SDV synthesizers use the UnixTimestampEncoder to convert datetimes into numerical values. Now, you can also extract the timezone so that your model can learn it as a separate feature. This is especially useful for learning and creating synthetic data that has a mix of different timezones.
Apply constraints when creating data from scratch. You can now apply any single-table constraint — including your own programmable constraint — to the DayZSynthesizer. This allows you to create random data from scratch, while still adhering to your business logic.