Tid: 18 april, kl 13- 14
Plats: B705

Abstract

This  talk will  provide an overview of the types of disclosure risks in traditional forms of statistical data products, how the risk and information loss can be quantified, and some recommended statistical disclosure control methods. Statistical agencies, however, are making more use of restricting and licensing data as a statistical disclosure control method  given increasing difficulties in producing safe data in an open environment. We will look at new strategies for data dissemination and discuss their potential and limitations. These include web-based applications where outputs are protected without the need for human intervention to check for disclosure risks, on-site data enclaves and its extension of connecting to remote servers and the production and release of 'synthetic' data where data is generated based on advanced statistical models. These new approaches have created a shift in  traditional types of disclosure risks that have formed the basis of statistical disclosure control methods up till now.  We define the notion of differential privacy as developed in the  Computer Science literature and how it relates to statistical disclosure control with  some applications.