The semantic discrimination rate metric for privacy measurements which questions the benefit of T-closeness over L-diversity

Louis-Philippe Sondeck (1, 2, 3, 4) , Maryline Laurent (5, 3, 1, 4) , Vincent Frey (2)

1 SAMOVAR - Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux
2 Orange Labs [Cesson-Sévigné]
3 TSP - RST - Département Réseaux et Services de Télécommunications
4 CNRS - Centre National de la Recherche Scientifique
5 R3S-SAMOVAR - Réseaux, Systèmes, Services, Sécurité

After a brief description of k-anonymity, l-diversity and t-closeness techniques, the paper presents the Discrimination Rate (DR) as a new metric based on information theory for measuring the privacy level of any anonymization technique. As far as we know, the DR is the first approach supporting fine grained privacy measurement down to attribute's values. Increased with the semantic dimension, the resulting semantic DR (SeDR) enables to: (1) tackle anonymity measurements from the attacker's perspective, (2) prove that t-closeness can give lower privacy protection than l-diversity

Mots clés: Anonymity metric; Semantic Discrimination rate; Discrimination Rate; Identifiability; K-anonymity; Tcloseness; L-diversity

Domaines: Base de données [cs.DB]; Cryptographie et sécurité [cs.CR]; Théorie de l'information [cs.IT]

Louis-Philippe Sondeck, Maryline Laurent, Vincent Frey. The semantic discrimination rate metric for privacy measurements which questions the benefit of T-closeness over L-diversity. SECRYPT 2017 : 14th International Conference on Security and Cryptography, Jul 2017, Madrid, Spain. pp.285 - 294, ⟨10.5220/0006418002850294⟩. ⟨hal-01576996⟩