Ascending partition method for differentially private histogram publication

摘要

Differential privacy has emerged as a promising scheme for protecting individual privacy in data release and data mining against adversaries with arbitrary background knowledge. It is often used to provide strong privacy guarantee in various types of data release, in particular, a histogram, which gives statistical summaries of the data set. Existing works on releasing sensitive data differentially with histogram mostly focus on boosting the accuracy and utility while preserving the individual privacy. This paper investigates how to increase the accuracy further.