Abstract
Viewing differential privacy through the operational lens of hypothesis testing, we derive tight data processing inequalities. Applications include optimal composition theorems that quantify the privacy degradation level during interactive querying and the optimality of the staircase privacy mechanism in several canonical settings.