Abstract
Popular approaches to differential privacy, such as the Laplace and exponential mechanisms, calibrate randomised smoothing through global sensitivity of the target non-private function. Bounding such sensitivity can be prohibitively complex for end users. As an alternative, we propose a straightforward sampler for estimating sensitivity of non-private mechanisms. Since our sensitivity estimates hold with high probability, any mechanism that would be (ϵ,δ)-differentially private under bounded global sensitivity automatically achieves (ϵ,δ,γ)-random differential privacy (Hall, Rinaldo, Wasserman, 2012), without any target-specific calculations required. We demonstrate on worked example learners how our usable approach adopts a naturally-relaxed privacy guarantee, while achieving more accurate releases even for non-private functions that are black-box computer programs.