Applied statistical disclosure control : methods and software

KeynoteThe demand for and volume of data from surveys, registers or other sources containing sensible information on persons or enterprises have been increased significantly over the last several years. At the same time, privacy protection principles and regulations have imposed restrictions on the access and use of individual data. Proper and secure microdata dissemination calls for the application of statistical disclosure control methods to data sets before release. Traditional approaches to (micro)data anonymization and statistical disclosure control include data perturbation methods, disclosure risk methods, methods to check the data utility of anonymized data sets. These traditional methods are enhanced by methods for simulating synthetic data sets using specialized model-based data methods of simulation and prediction that can deal with complex survey designs, missing values, hierarchical and cluster structures. One focus of the presentation is based on new developments and research for generating close-to-reality synthetic data sets using these special kinds of model-based approaches. The resulting data can work as a proxy of real-world data and they are useful for training purposes, agent-based and/or microsimulation experiments, remote execution as well as they can be provided as public-use files. The strength and weakness of the methods are highlighted and an (brief) application to the Euorpean Statistics of Income and Living Condition Survey is given by the use of R packages sdcMicro and simPop