Protecting Micro-Data Privacy: The Moment-Based Density Estimation Method and its Application

Privacy concerns pertaining to the release of confidential micro-level information are increasingly relevant to organisations and institutions. Controlling the dissemination of disclosure-prone micro-data by means of suppression, aggregation and perturbation techniques often entails different levels of effectiveness and drawbacks depending on the context and properties of the data. In this dissertation, we briefly review existing disclosure control methods for microdata and undertake a study demonstrating the applicability of micro-data methods to proportion data. This is achieved by using the sample size efficiency related to a simple hypothesis test for a fixed significance level and power, as a measure of statistical utility. We compare a query-based differential privacy mechanism to the multiplicative noise method for disclosure control and demonstrate that with the correct specification of noise parameters, the multiplicative noise method, which is a micro-data based method, achieves similar disclosure protection properties with reduced statistical efficiency costs.