Enhanced k-Anonymity model based on clustering to overcome Temporal attack in Privacy Preserving Data Publishing

Abstract

The infrastructure required for data storage and processing has become increasingly feasible, and hence, there has been a massive growth in the field of data acquisition and analysis. This acquired data is published, empowering organizations to make informed data-driven decisions based on previous trends. However, data publishing has led to the compromise of privacy as a result of the release of entity-specific information. Privacy-Preserving Data Publishing [1] can be accomplished by methods such as Data Swapping, Differential Privacy, and the likes of k-Anonymity. k-Anonymity is a well-established method used to protect the privacy of the data published. We propose a clustering-based novel algorithm named SAC or the Score, Arrange, and Cluster Algorithm to preserve privacy based on k-Anonymity. This method outperforms existing methods such as the Mondrian Algorithm by K. LeFevre and the One-pass K-means Algorithm by Jun-Lin Lin from a data quality perspective. SAC can be used to overcome temporal attack across subsequent releases of published data. To measure data quality post anonymization we present a metric that takes into account the relative loss in the information, that occurs while generalizing attribute values.