Top-down specialization for information and privacy preservation

By Benjamin C. M. Fung, Ke Wang, Philip S. Yu. IEEE International Conference on Data Engineering, 2005

686 citations recorded by the source corpus.

Releasing person-specific data in its most specific state poses a threat to individual privacy. This paper presents a practical and efficient algorithm for determining a generalized version of data that masks sensitive information and remains useful for modelling classification. The generalization of data is implemented by specializing or detailing the level of information in a top-down manner until a minimum privacy requirement is violated. This top-down specialization is natural and efficient for handling both categorical and continuous attributes. Our approach exploits the fact that data usually contains redundant structures for classification. While generalization may eliminate some structures, other structures emerge to help. Our results show that quality of classification can be preserved even for highly restrictive privacy requirements. This work has great applicability to both public and private sectors that share information for mutual benefits and productivity.

Subfields: Artificial Intelligence, Management Science and Operations Research, Sociology and Political Science

View Paper: https://www.semanticscholar.org/paper/fea4fab6bcf9872008735eed4cceaa54925ade41

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Management Science and Operations Research

In Sociology and Political Science

Citations in the corpus

Application of Local Differential Privacy to Collection of Indoor Positioning Data (2018, background): “Fung et al. presented the top-down specialization scheme in which the specialization process terminates if further specialization on quasi-identiﬁer attribute values violates k -anonymity requirement [8].”
Anatomy: simple and effective privacy preservation (2006, background): “RELATED WORK Generalization has been very well studied in the literature [1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16, 17, 18].”
On the tradeoff between privacy and utility in data publishing (2009, background): “Classification accuracy on the anonymized data has been evaluated in [18, 28, 10, 27, 5].”
Anonymizing Classification Data for Privacy Preservation (2007, methodology): “Greedy methods were proposed in [10], [11], [12], [13], [14], and [15].”
Mondrian Multidimensional K-Anonymity (2006, methodology): “There are a number of notions of k-anonymization quality [2, 6, 8, 10, 12, 13, 14, 15, 16], but intuitively, anonymization should generalize or perturb the data as little as is necessary to satisfy the k-anonymity constraint.”
A Survey of Quantification of Privacy Preserving Data Mining Algorithms (2008, background): “TheCM metric and the information gain privacy loss ratio [12] are m ore interesting measure of utility because it considers the possible application for the data.”
Deep Private-Feature Extraction (2018, background): “Many researches are presented to make a database k-anonymous [14], [15], [16], [17] and most of them are based on the generalization (e.”
Slicing: A New Approach to Privacy Preserving Data Publishing (2009, methodology): “We evaluate the quality of the anonymized data for classiﬁer learning, which has been used in [11, 18, 4].”
Incognito: efficient full-domain K-anonymity (2005, methodology): “Recently, top-down [7] and bottom-up [19] greedy heuristic algorithms were proposed for producing anonymous data that remains useful for building decision-tree classifiers.”
Personalized privacy preservation (2006, methodology): “Since the introduction of k-anonymity in [13, 15], numerous algorithms [5, 7, 8, 9, 10, 13, 15, 17] have been proposed to obtain kanonymous tables.”
A General Survey of Privacy-Preserving Data Mining Models and Algorithms (2008, methodology): “Therefore, even though the data is k -anonymized, the value of the sensitive attribute for that group of k records can be predicted exactly.”
Anonymity-preserving data collection (2005, background): “However, building an anonymous channel is a nontrivial task.”
Efficient k -Anonymization Using Clustering Techniques (2007, background): “In the hierarchy-based generalization [5, 11, 18, 29], a generalization-hierarchy is first defined for each quasiidentifier attribute.”
(α, k)-anonymity: an enhanced k-anonymity model for privacy preserving data publishing (2006, background): “One model isglobal recoding[11, 7, 5, 10] while the other is local recoding[11, 1].”
On static and dynamic methods for condensation-based privacy-preserving data mining (2008, methodology): “A number of papers have also appeared on the k -anonymity model [8; 7; 19; 17; 18; 23; 29; 38; 40] recently.”
Privacy-preserving data publishing: A survey of recent developments (2010, methodology): “Few works [Fung et al. 2005; 2007; Iyengar 2002; LeFevre et al. 2006b; Wang et al. 2004] have actually conducted such experiments, although many such as [Bayardo and Agrawal 2005] adopted CM in an attempt to address the classification problem.”
Minimality Attack in Privacy Preserving Data Publishing (2007, methodology): “Almost all known anonymization algorithms (including Incognito based methods [10, 12, 13, 11] and top-down approaches [6, 24, 22, 18]) try to reduce information loss of one form or another, and they all follow the above principle.”