Improved privacy-preserving Bayesian network parameter learning on vertically partitioned data

Zhiqiang Yang, Rebecca N. Wright

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

15 Scopus citations

Abstract

Privacy concerns often prevent different parties from sharing their data in order to carry out data mining applications on their joint data. Privacy-preserving data mining seeks to address this by enabling parties to jointly compute a data mining algorithm on distributed data without sharing their data. In this paper, we address a particular data mining problem, that of learning the parameters of Bayesian network on a vertically partitioned database. We provide a simple privacy-preserving protocol for learning the parameters of Bayesian network on vertically partitioned databases. In comparison to the previously known solution for this problem (Meng, Sivakumar, and Kargupta, 2004), our solution provides better performance, full privacy, and complete accuracy. In combination with our previous work on privacy-preserving learning of Bayesian network structure on vertically partitioned databases, this work provides a complete privacy-preserving protocol for learning Bayesian networks (both structure and parameters) on vertically partitioned data, with very little over-head beyond computing the structure alone.

Original languageEnglish (US)
Title of host publicationProceedings - International Workshop on Biomedical Data Engineering, BMDE2005
DOIs
StatePublished - 2005
Externally publishedYes
Event21st International Conference on Data Engineering Workshops 2005 - Tokyo, Japan
Duration: Apr 3 2005Apr 4 2005

Publication series

NameProceedings - International Workshop on Biomedical Data Engineering, BMDE2005
Volume2005

Other

Other21st International Conference on Data Engineering Workshops 2005
Country/TerritoryJapan
CityTokyo
Period4/3/054/4/05

ASJC Scopus subject areas

  • Engineering(all)

Fingerprint

Dive into the research topics of 'Improved privacy-preserving Bayesian network parameter learning on vertically partitioned data'. Together they form a unique fingerprint.

Cite this