Journal of Official Statistics, Vol.28, No.4, 2012. pp. 591–613
Confidentialising Exploratory Data Analysis Output in Remote Analysis
Christine M. OKeefe
Abstract:This article is concerned with the problem of balancing the competing objectives of allowing statistical analysis of confidential data while maintaining privacy and confidentiality. Traditional approaches to reducing the risk of disclosure typically involve modifying or confidentialising data before releasing it to users. In contrast, remote analysis enables analysts to submit statistical queries and receive output without direct access to data.In this article we discuss the implementation of remote analysis allowing exploratory data analysis on confidential data, where the system outputs are modified to protect confidentiality. To illustrate the effect of the modifications, we provide a comprehensive example comparing traditional and confidentialised output for a range of common exploratory data analyses on discrete and continuous data.We believe that confidentialised exploratory data analysis output is still useful, provided the analyst understands the confidentialisation process and its potential impact. Where the potential impact is judged to be too great, the analyst will need to seek another mode of access to the data.
Keywords:Confidentiality, privacy, remote access, remote data access, output checking
Copyright © Statistics Sweden 1996-2018. Open AccessISSN 0282-423XCreated and Maintained by OKS Group