Using Feature Selection to Improve the Utility of Differentially Private Data Publishing

Jafer, Yasser; Matwin, Stan; Sokolova, Marina

doi:10.1016/j.procs.2014.08.076

Cited by 17 publications

(8 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Domingo-Ferrer and Soria-Comas have shown that there is a theoretical relationship between -differential privacy and a stochastic extension of t-closeness and that satisfying t-closeness can imply -differential privacy under certain assumptions [8]. Moreover, Soria-Comas et al and Jafer et al have also combined k-anonymity and differential privacy [27,30]. While our approach uses k-anonymity in order to create a differentially private mechanism, these works employ k-anonymization to reduce the amount of noise that must be added.…”

Section: Related Workmentioning

confidence: 99%

SafePub: A Truthful Data Anonymization Algorithm With Strong Privacy Guarantees

Bild

Kuhn

Praßer

2018

Proceedings on Privacy Enhancing Technologies

View full text Add to dashboard Cite

Methods for privacy-preserving data publishing and analysis trade off privacy risks for individuals against the quality of output data. In this article, we present a data publishing algorithm that satisfies the differential privacy model. The transformations performed are truthful, which means that the algorithm does not perturb input data or generate synthetic output data. Instead, records are randomly drawn from the input dataset and the uniqueness of their features is reduced. This also offers an intuitive notion of privacy protection. Moreover, the approach is generic, as it can be parameterized with different objective functions to optimize its output towards different applications. We show this by integrating six well-known data quality models. We present an extensive analytical and experimental evaluation and a comparison with prior work. The results show that our algorithm is the first practical implementation of the described approach and that it can be used with reasonable privacy parameters resulting in high degrees of protection. Moreover, when parameterizing the generic method with an objective function quantifying the suitability of data for building statistical classifiers, we measured prediction accuracies that compare very well with results obtained using state-ofthe-art differentially private classification algorithms.

show abstract

Section: Related Workmentioning

confidence: 99%

SafePub: A Truthful Data Anonymization Algorithm With Strong Privacy Guarantees

Bild

Kuhn

Praßer

2018

Proceedings on Privacy Enhancing Technologies

View full text Add to dashboard Cite

show abstract

“…The specific process of how to calculate the error lower bound and function sensitivity was given. In 2014, Jafer et al [26] proposed the TOP Diff algorithm, which can reduce the operation time of anonymization of large data sets by k-anonymity and differential privacy after feature extraction of original data sets, and obtain higher data utility by mutually adjusting the k value of the anonymization level and the privacy budget. The algorithm can only be applied in non-numerical data.…”

Section: Related Workmentioning

confidence: 99%

Differential identifiability clustering algorithms for big data analysis

Shang

Zhao

Ren

et al. 2021

Sci. China Inf. Sci.

View full text Add to dashboard Cite

Individual privacy preservation has become an important issue with the development of big data technology. The definition of ρ-differential identifiability (DI) precisely matches the legal definitions of privacy, which can provide an easy parameterization approach for practitioners so that they can set privacy parameters based on the privacy concept of individual identifiability. However, differential identifiability is currently only applied to some simple queries and achieved by Laplace mechanism, which cannot satisfy complex privacy preservation issues in big data analysis. In this paper, we propose a new exponential mechanism and composition properties of differential identifiability, and then apply differential identifiability to k-means and k-prototypes algorithms on MapReduce framework. DI k-means algorithm uses the usual Laplace mechanism and composition properties for numerical databases, while DI k-prototypes algorithm uses the new exponential mechanism and composition properties for mixed databases. The experimental results show that both DI k-means and DI k-prototypes algorithms satisfy differential identifiability.

show abstract

“…It assigns a generalization penalty over the rows of the dataset and uses a genetic algorithm for the classification task, but for a fair comparison we use CM criterion in the Greedy algorithm and with the selected features we identical setup for classification. Laplace-DP [21] is a method to use feature selection for ǫ-differential private data publishing. Authors in [21] utilize Laplace mechanism [13] for ǫ-differential privacy guarantee.…”

Section: Experimental Settingmentioning

confidence: 99%

“…Laplace-DP [21] is a method to use feature selection for ǫ-differential private data publishing. Authors in [21] utilize Laplace mechanism [13] for ǫ-differential privacy guarantee. To compare with their method, we first compute the utility of each feature x i ∈ I as its true output using HamDist function in Definition 8 denoted as H(x i ).…”

Section: Experimental Settingmentioning

confidence: 99%

See 1 more Smart Citation

Feature Selection for Classification under Anonymity Constraint

Zhang¹,

Mohammed²,

Dave³

et al. 2015

Preprint

View full text Add to dashboard Cite

Over the last decade, proliferation of various online platforms and their increasing adoption by billions of users have heightened the privacy risk of a user enormously. In fact, security researchers have shown that sparse microdata containing information about online activities of a user although anonymous, can still be used to disclose the identity of the user by cross-referencing the data with other data sources. To preserve the privacy of a user, in existing works several methods (k-anonymity, ℓ-diversity, differential privacy) are proposed for ensuring that a dataset bears small identity disclosure risk. However, the majority of these methods modify the data in isolation, without considering their utility in subsequent knowledge discovery tasks, which makes these datasets less informative. In this work, we consider labeled data that are generally used for classification, and propose two methods for feature selection considering two goals: first, on the reduced feature set the data has small disclosure risk, and second, the utility of the data is preserved for performing a classification task. Experimental results on various real-world datasets show that the method is effective and useful in practice.

show abstract

Using Feature Selection to Improve the Utility of Differentially Private Data Publishing

Cited by 17 publications

References 13 publications

SafePub: A Truthful Data Anonymization Algorithm With Strong Privacy Guarantees

SafePub: A Truthful Data Anonymization Algorithm With Strong Privacy Guarantees

Differential identifiability clustering algorithms for big data analysis

Feature Selection for Classification under Anonymity Constraint

Contact Info

Product

Resources

About