Package org.deidentifier.arx.criteria
package org.deidentifier.arx.criteria
This package implements different variants of class-based privacy criteria,
such as k-anonymity, l-diversity, t-closeness and d-presence.
Moreover, this package implements sample-based criteria, such as thresholds on the average re-identification risk, population uniqueness and sample uniqueness.
k-anonymity and d-presence and the sample-based criteria are implicit privacy criteria, i.e., they are implicitly bound to the quasi-identifiers, while the other criteria are explicitly bound to a specific sensitive attribute.
Moreover, this package implements sample-based criteria, such as thresholds on the average re-identification risk, population uniqueness and sample uniqueness.
k-anonymity and d-presence and the sample-based criteria are implicit privacy criteria, i.e., they are implicitly bound to the quasi-identifiers, while the other criteria are explicitly bound to a specific sensitive attribute.
-
KlasseBeschreibungThis criterion ensures that an estimate for the average re-identification risk falls below a given threshold.Basic-beta-Likeness:
Jianneng Cao, Panagiotis Karras:
Publishing Microdata with a Robust Privacy Guarantee
VLDB 2012.Delta-disclosure privacy as proposed in:
Justin Brickell and Vitaly Shmatikov:
The Cost of Privacy: Destruction of Data-mining Utility in Anonymized Data Publishing
Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
2008The distinct l-diversity privacy criterion.The d-presence criterion Published in: Nergiz M, Atzori M, Clifton C.(e,d)-Differential Privacy implemented with SafePub as proposed in: Bild R, Kuhn KA, Prasser F.Enhanced-beta-Likeness:
Jianneng Cao, Panagiotis Karras:
Publishing Microdata with a Robust Privacy Guarantee
VLDB 2012.The entropy l-diversity privacy model.Enumerator of entropy estimators for the entropy-l-diversity privacy model.The t-closeness criterion with equal-distance EMD.A privacy criterion that is explicitly bound to a sensitive attribute.The t-closeness criterion with hierarchical-distance EMD.A privacy criterion that is implicitly bound to the quasi-identifiers.This is a special criterion that does not enforce any privacy guarantees but allows to define a data subset.The k-anonymity criterion Published in: Sweeney L.This class implements the k-map privacy model as proposed by Latanya Sweeney.
As an alternative to explicitly providing data about the underlying population, cell sizes can be can be estimated with the D3 (Poisson) and D4 (zero-truncated Poisson) estimators proposed in:
K.Estimators for cell sizes in the population.An abstract base class for l-diversity criteria Published in: Machanavajjhala A, Kifer D, Gehrke J.The t-closeness criterion for ordered attributes.This criterion ensures that the population uniqueness falls below a given threshold.An abstract base class for privacy criteria.Privacy model for the game theoretic approach proposed in: A Game Theoretic Framework for Analyzing Re-Identification Risk.Privacy model for the "no-attack" variant of the game theoretic approach proposed in: A Game Theoretic Framework for Analyzing Re-Identification Risk.Privacy model for the game theoretic approach proposed in: A Game Theoretic Framework for Analyzing Re-Identification Risk.Privacy model for the "no-attack" variant of the game theoretic approach proposed in: A Game Theoretic Framework for Analyzing Re-Identification Risk.The recursive-(c,l)-diversity criterion.Abstract class for criteria that ensure that a certain risk measure is lower than or equal to a given thresholdAn abstract base class for sample-based privacy criteria.This criterion ensures that the sample uniqueness falls below a given threshold.An abstract base class for t-closeness criteria as proposed in: Li N, Li T, Venkatasubramanian S.