Disclosure detection in research environments in practice

Ritchie, Felix

Abstract

There is an increasing demand for access to raw confidential data, and NSIs have
responded by setting up controlled research facilities. However, the most common approaches to
statistical disclosure detection and control (SDDC) struggle to accommodate the infinite variety of
outputs produced in research environments. The main problems are designing statistical disclosure
control (SDC) rules for unknown transformations of the data, and in managing the potential volume of
outputs needing review.
Research facilities need a different approach to SDDC. In the UK, ONS has developed an approach
based around classes of output, where the time devoted to checking outputs can be concentrated on the
more unsafe outputs.
Defining "safe" and "unsafe" outputs based on the functional form of the model improves the
efficiency and security of confidentiality checking, but is not straightforward. This paper outlines the
broad approach, and then takes specific examples to show how the UK rules on analytical outputs
(and the conditions attached to them) have been developed.

• Presentation Conference Type: Conference Paper (unpublished)
• Conference Name: UNECE/Eurostat work session on statistical data confidentiality - 2007
• Start Date: Dec 17, 2007
• End Date: Dec 19, 2007
• Publication Date: Dec 17, 2007
• Publicly Available Date: Jun 8, 2019
• Peer Reviewed: Not Peer Reviewed
• Pages: 399-406
• Keywords: disclosure, detection, research, environments
• Public URL: https://uwe-repository.worktribe.com/output/1023051
• Publisher URL: http://epp.eurostat.ec.europa.eu/portal/page/portal/conferences/documents/unece_es_work_session_statistical_data_conf/TOPIC%203-WP.37%20SP%20RITCHIE.PDF
• Related Public URLs: http://dx.doi.org/10.2901/Eurostat.C2007.004
• Additional Information: Title of Conference or Conference Proceedings : UNECE/Eurostat work session on statistical data confidentiality - 2007