Tree traversal to achieve generalization for data de-identification
Jose Carlos Donaldo Hernandez, David Anastacio Garcia, Farshad Rabbani · International journal of open information technologies · 2018
Every day data is published of different types and from various sources. Data de-identification protects the privacy of most of this data before its publication. Over the recent years, a technique proposed by Dr. Sweeney, known as kanonymization as a means for privacy protection has gained great popularity. There has been intensive research involving this method and many alterations, in the hope to find an optimal solution in real-time to the generalization problem. To achieve either generalization or suppression, researchers have used different types of heuristics, most of them being tree-based. Although this is a heavily investigated area, there is no simple method to prepare data for generalization; in theory, there are infinite methods for data preparation and partitioning. In this research, we first propose the use of commonly known algorithms to prepare data and achieve generalization. We then introduce the use of tree-based algorithms and tree traversal as the mechanism to achieve data generalization. We further investigate them, by comparing the quality of generalization sets obtained in each traversal method, in the hope to determine which method is best.