Keynote - Privacy & Data Protection : Policy and Business Trends vi Harriet P . Pearson Towards Application - Oriented Data

Data anonymization is of increasing importance for allowing sharing of individual data for a variety of data analysis and mining applications. Most of existing work on data anonymization optimizes the anonymization in terms of data utility typically through one-size-fits-all measures such as data discernibility. Our primary viewpoint in this paper is that each target application may have a unique need of the data and the best way of measuring data utility is based on the analysis task for which the anonymized data will ultimately be used. We take a top-down analysis of typical application scenarios and derive applicationoriented anonymization criteria. We propose a prioritized anonymization scheme where we prioritize the attributes for anonymization based on how important and critical they are to the application needs. Finally, we present preliminary results that show the benefits of our approach.