Fully Adaptive Composition for Gaussian Differential Privacy
This addresses a key challenge in differential privacy for adaptive data analysis, providing a theoretical foundation for privacy-preserving algorithms in sensitive data settings.
The paper tackles the problem of maintaining privacy guarantees under fully adaptive composition in Gaussian Differential Privacy, showing that it composes gracefully with the same parameters as nonadaptive composition.
We show that Gaussian Differential Privacy, a variant of differential privacy tailored to the analysis of Gaussian noise addition, composes gracefully even in the presence of a fully adaptive analyst. Such an analyst selects mechanisms (to be run on a sensitive data set) and their privacy budgets adaptively, that is, based on the answers from other mechanisms run previously on the same data set. In the language of Rogers, Roth, Ullman and Vadhan, this gives a filter for GDP with the same parameters as for nonadaptive composition.