AIM: A New Privacy Preservation Algorithm for Incomplete Microdata Based on Anatomy

摘要

Although many algorithms have been developed to achieve privacy preserving data publishing, few of them can handle incomplete microdata. In this paper, we first show that traditional algorithms based on suppression and generalization cause huge information loss on incomplete microdata. Then, we propose AIM (anatomy for incomplete microdata), a linear-time algorithm based on anatomy, aiming to retain more information in incomplete microdata. Different from previous algorithms, AIM treats missing values as normal value, which greatly reduce the number of records being suppressed. Compared to anatomy, AIM supports more kinds of datasets, by employing a new residue-assignment mechanism, and is applicable to all privacy principles. Results of extensive experiments based on real datasets show that AIM provides highly accurate aggregate information for the incomplete microdata.