Utility-based anonymization using local recoding

Abstract

Privacy becomes a more and more serious concern in applications involving microdata. Recently, efficient anonymization has attracted much research work. Most of the previous methods use global recoding, which maps the domains of the quasi-identifier attributes to generalized or changed values. However, global recoding may not always achieve effective anonymization in terms of discernability and query answering accuracy using the anonymized data. Moreover, anonymized data is often for analysis. As well accepted in many analytical applications, different attributes in a data set may have different utility in the analysis. The utility of attributes has not been considered in the previous methods.In this paper, we study the problem of utility-based anonymization. First, we propose a simple framework to specify utility of attributes. The framework covers both numeric and categorical data. Second, we develop two simple yet efficient heuristic local recoding methods for utility-based anonymization. Our extensive performance study using both real data sets and synthetic data sets shows that our methods outperform the state-of-the-art multidimensional global recoding methods in both discernability and query answering accuracy. Furthermore, our utility-based method can boost the quality of analysis using the anonymized data.

Utility Enhancement for Privacy Preserving Health Data Publishing

“…To capture data utility, some criteria measure the utility loss that is incurred by generalization based on generalization hierarchies, such as Discernability Measure (DM) [1], Utility Measure (UM) [17], Relative Error (RE) [18], Normalized Certainty Penalty (NCP) [16] etc. DM and RE is calculated based on the number of generalized group and suppressed group that overlap with the original data.…”

“…A subspace that contains at least k points forms a k-anonymous group [18]. The main idea of clustering-based anonymization is to create clusters containing at least k records in each cluster separately [16]. Fung et al [19] presented an effective top-down approach by introducing multiple virtual identifiers for utilizing information and privacy-guided specialization.…”

Personalized Privacy Preserving Publication of Transactional Datasets Using Concept Learning

“…Even if the identifying attributes like name is removed, an attacker may be able to associate records with specific persons using combinations of other attributes (e.g., Postal Code; Gender; birth-date), called quasi-identifiers (QID suppression [5]. Generalization replaces their actual QID values with more general ones.…”

“…The Classification Metric (CM) [10] is suitable when the purpose of the anonymized data is to train a classifier, whereas the Discernibility Metric (DM) [9] measures the cardinality of the anonymized groups. More accurate is the Generalized Loss Metric [10] and the similar Normalized Certainty Penalty (NCP) [5]. In the case of categorical attributes NCP is defined with respect to the hierarchy.…”

“…In the case of categorical attributes NCP is defined with respect to the hierarchy. On the similar guidelines, the information loss is defined in (5). Let x be an item in T. Then:…”