Proportional Differential Privacy (PDP): A new Approach for Differentially Private Histogram Release based on Buckets Densities
Mohammed Aissaoui · 2020
Differential privacy is an effective model for releasing the results of statistical queries on sensitive data. It can be used to release different types of data, in particular, histograms, which provide a useful summary of a dataset. Several differentially private histogram releasing schemes have recently been proposed. However, these methods impose the same amount of privacy control over all data, by using one private budgeting value. In this paper, we introduce a new approach called Proportional Differential Privacy (PDP), based on the density of buckets. The aim is to use different levels of privacy in order to publish Differential Privacy histograms having a reduced queries' error. Carried out experiments on real-life datasets highlights very encouraging results in terms of information loss decrease.