Searching for better randomized response schemes for privacy-preserving data mining

Zhengli Huang, Wenliang Du, Zhouxuan Teng

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations

Abstract

To preserve user privacy in Privacy-Preserving Data Mining (PPDM), the randomized response (RR) technique is widely used for categorical data. Although various RR schemes have been proposed, there is no study to systematically compare them in order to find optimal RR schemes. In the paper, we choose the R-U (Risk-Utility) confidentiality map to compare different randomization schemes. Using the R-U map as our metric, we present an optimal RR scheme for binary data, which helps us find an optimal class of RR matrices. From this optimal scheme, we have discovered several heuristic rules among the elements in the optimal class. We generalize these rules to find optimal class of RR matrices for categorical data. Based on these rules, we propose an RR scheme to find a class of RR matrices for categorical data. Our experimental results have shown that our scheme has much better performance than the existing RR schemes.

Original languageEnglish (US)
Title of host publicationKnowledge Discovery in Database
Subtitle of host publicationPKDD 2007 - 11th European Conference on Principles and Practice of Knowledge Discovery in Databases, Proceedings
Pages487-497
Number of pages11
StatePublished - Dec 1 2007
Event11th European Conference on Principles and Practice of Knowledge Discovery in Databases, PKDD 2007 - Warsaw, Poland
Duration: Sep 17 2007Sep 21 2007

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4702 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other11th European Conference on Principles and Practice of Knowledge Discovery in Databases, PKDD 2007
CountryPoland
CityWarsaw
Period9/17/079/21/07

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Searching for better randomized response schemes for privacy-preserving data mining'. Together they form a unique fingerprint.

  • Cite this

    Huang, Z., Du, W., & Teng, Z. (2007). Searching for better randomized response schemes for privacy-preserving data mining. In Knowledge Discovery in Database: PKDD 2007 - 11th European Conference on Principles and Practice of Knowledge Discovery in Databases, Proceedings (pp. 487-497). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4702 LNAI).