Co-clustering for differentially private synthetic data generation - Archive ouverte HAL Accéder directement au contenu
Chapitre D'ouvrage Année : 2017

Co-clustering for differentially private synthetic data generation

Tarek Benkhelif
  • Fonction : Auteur
Françoise Fessant
  • Fonction : Auteur
  • PersonId : 856215

Résumé

We propose a methodology to anonymize microdata (i.e. a table of n individuals described by d attributes). The goal is to be able to release an anonymized data table built from the original data while meeting the differential privacy requirements. The proposed solution combines co-clustering with synthetic data generation to produce anonymized data. First, a data independent partitioning on the domains is used to generate a perturbed multidimensional histogram; a multidi-mensional co-clustering is then performed on the noisy histogram resulting in a partitioning scheme. This differentially private co-clustering phase aims to form attribute values clusters and thus, limits the impact of the noise addition in the second phase. Finally, the obtained scheme is used to partition the original data in a differentially private fashion. Synthetic individuals can then be drawn from the partitions. We show through experiments that our solution outperforms existing approaches and we demonstrate that the produced synthetic data preserve sufficient information and can be used for several datamining tasks.
Fichier principal
Vignette du fichier
PAP2017_Revised_Book.pdf (964.47 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01664224 , version 1 (14-12-2017)

Identifiants

  • HAL Id : hal-01664224 , version 1

Citer

Tarek Benkhelif, Françoise Fessant, Fabrice Clérot, Guillaume Raschia. Co-clustering for differentially private synthetic data generation. Personal Analytics and Privacy. An Individual and Collective Perspective, 2017. ⟨hal-01664224⟩
127 Consultations
170 Téléchargements

Partager

Gmail Facebook X LinkedIn More