Efficient Table Anonymization for Aggregate Query Answering

Privacy protection is a major concern when microdata is released for ad hoc analyses. Anonymization schemes have to guarantee privacy goals, as well as preserve sufficient information to support reasonably accurate answers to ad hoc queries. In this paper, we focus on the case when the sensitive attributes are numerical (e.g., salary) for which (k,e)-anonymity was shown to be an appropriate privacy goal. We develop efficient algorithms for two optimization criteria for (k,e)-anonymity schemes, significantly improving on previous results. We evaluate our methods on a large real dataset, and show that they are scalable and accurate.