Anonymization methods for taxonomic microdata

  • Authors:

  • Josep Domingo-Ferrer;Krish Muralidhar;Guillem Rufian-Torrell

  • Affiliations:

  • Department of Computer Engineering and Maths, Universitat Rovira i Virgili, UNESCO Chair in Data Privacy, Tarragona, Catalonia, Spain;Gatton College of Business & Economics, University of Kentucky, Lexington, KY;Department of Computer Engineering and Maths, Universitat Rovira i Virgili, UNESCO Chair in Data Privacy, Tarragona, Catalonia, Spain

  • Venue:
  • PSD'12 Proceedings of the 2012 international conference on Privacy in Statistical Databases
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Often microdata sets contain attributes which are neither numerical nor ordinal, but take nominal values from a taxonomy, ontology or classification (e.g. diagnosis in a medical data set about patients, economic activity in an economic data set, etc.). Such data sets must be anonymized if transferred outside the data collector's premises (e.g. hospital or national statistical office), say, for research purposes. The literature on microdata anonymization methods is relatively limited for nominal data. Multiple imputation is a usual choice for such data, but it has computational problems when nominal attributes can take many possible different values. In this paper, we provide anonymization methods for data sets which include nominal taxonomic attributes with many possible different values. We show how to adapt to the case of taxonomic attributes two anonymization methods, data shuffling and microaggregation, that were originally designed for numerical attributes. The above adaptation relies on a hierarchy-aware numerical mapping of nominal categories, which we call marginality. The resulting adapted methods circumvent the computational problems of multiple imputation and take the semantics of the taxonomy into account.