Skip to main navigation Skip to search Skip to main content

A privacy protection method for learning artificial neural network on vertically distributed data

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

For mining privacy data that can not be seen directly, the privacy preserving data mining (PPDM) is needed. As far as we know, for neural network learning on vertically distributed databases, there is no good enough PPDM method. For solving it, a privacy preserving method for learning neural networks on vertically distributed data is proposed by this paper. This method designs protocols to exchange essential information for learning neural networks without opening private data. The learning results with this proposed method are the same as the results with the original BP algorithm without considering privacy preservation. And, in the learning process, every node cannot get the details of other nodes’ data.

Original languageEnglish
Title of host publicationRecent Developments in Mechatronics and Intelligent Robotics - Proceedings of International Conference on Mechatronics and Intelligent Robotics ICMIR2018
EditorsJohn Wang, Kevin Deng, Srikanta Patnaik, Zhengtao Yu
PublisherSpringer Verlag
Pages1159-1167
Number of pages9
ISBN (Print)9783030002138
DOIs
StatePublished - 2019
Externally publishedYes
EventInternational Conference on Mechatronics and Intelligent Robotics, ICMIR 2018 - Kunming, China
Duration: 19 May 201820 May 2018

Publication series

NameAdvances in Intelligent Systems and Computing
Volume856
ISSN (Print)2194-5357

Conference

ConferenceInternational Conference on Mechatronics and Intelligent Robotics, ICMIR 2018
Country/TerritoryChina
CityKunming
Period19/05/1820/05/18

Keywords

  • Neural network
  • Privacy preserving
  • Secure multi-party computation
  • Vertically distributed databases

Fingerprint

Dive into the research topics of 'A privacy protection method for learning artificial neural network on vertically distributed data'. Together they form a unique fingerprint.

Cite this