Encoding of numerical data for privacy-preserving record linkage

Lea Demelius*, Karl Kreiner, Dieter Hayn, Michael Nitzlnader, Günter Schreier

*Corresponding author for this work

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


Background: Privacy-preserving record linkage (PPRL) is the process of detecting dataset entries that refer to the same individual within two independent datasets, without disclosing any personal information. While applied in different fields, it particularly attained importance in the medical sector. One popular PPRL method are Bloom filters. However, Bloom filters were originally used for encoding strings only. Objectives: This paper evaluates an encoding method specifically designed for numerical data and adjusts it for encoding geocoordinates in Bloom filters. Methods: The proposed numerical encoding of geocoordinates is compared to the string-based method by using synthetic data. Results: The proposed method for encoding geocoordinates in Bloom filters attains a higher recall and precision than the conventional string encoding. Conclusion: Numerical encoding has the potential of increasing the record linkage quality of Bloom filters, as well as their privacy level.

Original languageEnglish
Title of host publicationdHealth 2020 - Biomedical Informatics for Health and Care
Subtitle of host publicationProceedings of the 14th Health Informatics Meets Digital Health Conference
EditorsGunter Schreier, Dieter Hayn, Alphons Eggerth
PublisherIOS Press BV
Number of pages8
ISBN (Electronic)9781643680842
Publication statusPublished - 23 Jun 2020

Publication series

NameStudies in Health Technology and Informatics
ISSN (Print)0926-9630
ISSN (Electronic)1879-8365


  • Algorithms
  • Computerized
  • Confidentiality
  • Medical record linkage
  • Medical records system
  • Personally identifiable information
  • Privacy

ASJC Scopus subject areas

  • Biomedical Engineering
  • Health Informatics
  • Health Information Management


Dive into the research topics of 'Encoding of numerical data for privacy-preserving record linkage'. Together they form a unique fingerprint.

Cite this