Abstract:Differential privacy is a cryptographically-motivated approach to privacy
that has become a very active field of research over the last decade in
theoretical computer science and machine learning. In this paradigm one assumes
there is a trusted curator who holds the data of individuals in a database and
the goal of privacy is to simultaneously protect individual data while allowing
the release of global characteristics of the database. In this setting we
introduce a general framework for parametric infe ...
Can a differentially private estimator be characterized via its influence function?
Medina shows that bounded influence estimators can be naturally used to construct differentially private estimators. However, he also shows that there are differentially private estimators which do not have an asymptotically bounded influence function.
Is there a condition on the influence function of an estimator such that medina's associated estimator or some other natural associated estimator will be differentially private if and only if it satisfies this condition?