What are the privacy protection measures we take when generating synthetic data?
Mainly, those are metrics to prevent overfitting, looking at distance-measures. This means they check how close the synthetic data is to the original data. If that gets too close, there might be a privacy risk. These metrics make sure that the synthetics data does not get too close to the original data. Additionally, when doing this, the Syntho Engine also uses a holdout set to be able to do this in a fair way.