Attribute Utility Motivated k-anonymization of datasets to support the heterogeneous needs of biomedical researchers

AMIA Annu Symp Proc. 2011:2011:1573-82. Epub 2011 Oct 22.

Abstract

In order to support the increasing need to share electronic health data for research purposes, various methods have been proposed for privacy preservation including k-anonymity. Many k-anonymity models provide the same level of anoymization regardless of practical need, which may decrease the utility of the dataset for a particular research study. In this study, we explore extensions to the k-anonymity algorithm that aim to satisfy the heterogeneous needs of different researchers while preserving privacy as well as utility of the dataset. The proposed algorithm, Attribute Utility Motivated k-anonymization (AUM), involves analyzing the characteristics of attributes and utilizing them to minimize information loss during the anonymization process. Through comparison with two existing algorithms, Mondrian and Incognito, preliminary results indicate that AUM may preserve more information from original datasets thus providing higher quality results with lower distortion.

MeSH terms

  • Algorithms*
  • Biomedical Research*
  • Computer Security
  • Confidentiality*
  • Health Insurance Portability and Accountability Act
  • Medical Records Systems, Computerized*
  • United States