A Survey of Inference Control Methods for Privacy-Preserving Data Mining

Abstract

Inference control in databases, also known as Statistical Disclosure Control (SDC), is about protecting data so they can be published without revealing confidential information that can be linked to specific individuals among those to which the data correspond. This is an important application in several areas, such as official statistics, health statistics, e-commerce (sharing of consumer data), etc. Since data protection ultimately means data modification, the challenge for SDC is to achieve protection with minimum loss of the accuracy sought by database users. In this chapter, we survey the current state of the art in SDC methods for protecting individual data (microdata). We discuss several information loss and disclosure risk measures and analyze several ways of combining them to assess the performance of the various methods. Last but not least, topics which need more research in the area are identified and possible directions hinted.

Data Anonymity Meets Non-discrimination

“…A common tool for statistical analysis is provided by a 1 The use of PD (resp., PND) attributes in decision making does not necessarily lead to (or exclude) discriminatory decisions [5], [7]. This motivates the adjective "potentially".…”

“…Several inference control methods have been proposed in privacy-preserving data mining for protecting micro-data from the risk of revealing confidential information, such as identities and sensitive attribute values [1]. Ultimately, private data protection consists of data transformations, such as perturbations, generalizations, or suppressions, that achieve a measurable level of privacy, according to some formal model, such as k-anonymity [2] or t-closeness [3].…”

“…If no such cut exists for P, then the values of every PND attribute are replaced by the range "[min, max]" of such an attribute in P. This substitution is performed by the PND ranges() function in Alg. 1. Differently from what would occur by a direct application of the Mondrian algorithm, we prevent cutting on the PD attribute, which in Thms.…”

A Comprehensive Survey on Privacy Preserving Big Data Mining

“…The attributes can be classified as [16] identifiers, quasi-identifiers, confidential outcome attributes and non-confidential outcome attributes. There are several approaches implemented for privacy preserving data mining.…”

Anti-discrimination Analysis Using Privacy Attack Strategies

“…We follow the intriguing parallel between the role of the anti-discrimination authority in discrimination data analysis and the role of an attacker in privacypreserving data publishing [1,4,5] -an unauthorized (possibly malicious) entity. Several attack strategies have been proposed in the literature, which model the reasonings of an attacker and its background knowledge.…”