Statistical disclosure control : protection of confidential longitudinal data

Abstract

National Statistical Offices(NSOs) collect high quality data in order to disseminate them to researchers and trusted organizations, they must first ensure that the respondents’ confidentiality is protected, as Statistic Offices are required by either their respective national law or the General Data Protection Regulation (”GDPR”) at EU level to fulfill that. That is when statistical disclosure methods come into play. Statistical disclosure control (SDC) is the general term describing various anonymization techniques which are applied to microdata. The biggest challenges the staff of National Statistical Offices are faced with when applying masking methods is the trade-off between information loss and disclosure risk. The safer the data after anonymization process the higher the information loss, which could even make the data unusable for researchers who need to perform further statistical analysis with data, as for example analysis in regards to socio-economic topics. Often data collected for socio-economic aspects, such as household surveys, for example the EU-SILC at EU level, are of longitudinal form. While the process of statistical disclosure control became an indispensable part of the data dissemination process, the application of anonymization techniques to longitudinal data, as a special case to of traditional microdata, is not a much touched topic in recent literature. Therefor, the main focus of this thesis is to carry out algorithms for the anonymization of longitudinal data and afterward examine the trade-off between information loss and data utility.