Differentially private data release: Improving utility with wavelets and bayesian networks

Abstract

Privacy-preserving data publishing is an important problem that has been the focus of extensive study. The state-of-the-art privacy model for this problem is differential privacy, which offers a strong degree of privacy protection without making restrictive assumptions about the adversary. In this paper, we review two methods, Privelet and PrivBayes, for improving utility in differentially private data publishing. Privelet utilizes wavelet transforms to ensure that any range-count query can be answered with noise variance that is polylogarithmic to the size of the input data domain. Meanwhile, PrivBayes employs Bayesian networks to publish high-dimensional datasets without incurring prohibitive computation overheads or excessive noise injection. © 2014 Springer International Publishing Switzerland.