A practical approximation algorithm for optimal k-anonymity

作者：Kenig Batya; Tassa Tamir^*

来源：Data Mining and Knowledge Discovery, 2012, 25(1): 134-168.

DOI：10.1007/s10618-011-0235-9

• Privacy-preserving data mining
• k-Anonymity
• l-Diversity
• Approximation algorithms for NP-hard problems
• Frequent generalized itemsets

k-Anonymity is a privacy preserving method for limiting disclosure of private information in data mining. The process of anonymizing a database table typically involves generalizing table entries and, consequently, it incurs loss of relevant information. This motivates the search for anonymization algorithms that achieve the required level of anonymization while incurring a minimal loss of information. The problem of k-anonymization with minimal loss of information is NP-hard. We present a practical approximation algorithm that enables solving the k-anonymization problem with an approximation guarantee of O(ln k). That algorithm improves an algorithm due to Aggarwal et al. (Proceedings of the international conference on database theory (ICDT), 2005) that offers an approximation guarantee of O(k), and generalizes that of Park and Shim (SIGMOD '07: proceedings of the 2007 ACM SIGMOD international conference on management of data, 2007) that was limited to the case of generalization by suppression. Our algorithm uses techniques that we introduce herein for mining closed frequent generalized records. Our experiments show that the significance of our algorithm is not limited only to the theory of k-anonymization. The proposed algorithm achieves lower information losses than the leading approximation algorithm, as well as the leading heuristic algorithms. A modified version of our algorithm that issues a""-diverse k-anonymizations also achieves lower information losses than the corresponding modified versions of the leading algorithms.

• 出版日期2012-7