Encoding Hierarchical Classification Codes for Privacy-Preserving Record Linkage Using Bloom Filters

Abstract

Hierarchical classification codes are widely used in many scientific fields. Such codes might reveal sensitive personal information, for example medical conditions or occupations. This paper introduces a new encoding technique for encrypting sensitive codes, which preserves the hierarchical similarity of the codes. The encoding was developed for the use of hierarchical codes in Privacy-preserving Record Linkage (PPRL). The technique is demonstrated with real-world survey data containing occupational codes (ISCO codes). After describing the construction and its similarity preserving properties, Hierarchy Preserving Bloom Filters (HPBF) are compared with positional q-grams and standard Bloom filters in a PPRL context. The method presented here is similarity preserving for hierarchies, privacy-preserving and will increase linkage quality when used in Bloom filter-based PPRL.