Techniques & safeguards
Differential privacy
DP
Standards, techniques & other
A mathematical guarantee that the result of an analysis, or a trained model, barely changes whether or not any one person's data was included, achieved by adding calibrated noise. The privacy budget (epsilon) sets the trade-off with accuracy.