A K-Anonymizing Approach for Preventing Link Attacks in Data Publishing

摘要

K-anonymization is an important approach to protect data privacy in data publishing. It is desired to publish k-anomymized data with less information loss. However, the existing algorithms are not feasible enough to satisfy such a requirement. We propose a k-anonymization approach, Classfly for publishing as much data as possible. For any attribute, in stead of generalizing all values, Classfly only generalizes partial values that do not satisfy fc-anonymization. As a side-effect, Classfly provides higher efficiency than existing approaches, since not all data need to be generalized. Classfly also considers the case of satisfying multiple anonymity constraints in one published table, which makes it more feasible for real applications. Experimental results show that the proposed Classfly approach can efficiently generate a published table with less information loss.