Year: 2008
Type: article
Abstract: The identity of patients must be protected when patient data are shared. The two most commonly used models to protect identity of patients are L-diversity and K-anonymity. However, existing work mainly considers data sets with a single sensitive attribute, while patient data often contain multiple sensitive attributes (e.g., diagnosis and treatment). This article shows that although the K-anonymity model can be trivially extended to multiple sensitive attributes, the L-diversity model cannot. The reason is that achieving L-diversity for each individual sensitive attribute does not guarantee L-diversity over all sensitive attributes. We propose a new model that extends L-diversity and K-anonymity to multiple sensitive attributes and propose a practical method to implement this model. Experimental results demonstrate the effectiveness of our approach.
Authors: Tamas S. Gal, Zhiyuan Chen, Aryya Gangopadhyay
Institution: University of Maryland, Baltimore County
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W2165111576
