Reference Values Based Hardening for Bloom Filters Based Privacy-Preserving Record Linkage

Abstract

Privacy-preserving record linkage (PPRL) is the process of identifying records that refer to the same entities across different data-bases without revealing any sensitive information about these entities. A popular PPRL technique that is efficient and effective is Bloom filter encoding. However, recent research has shown that Bloom filters are vulnerable to cryptanalysis attacks that aim to re-identify sensitive attribute values encoded into Bloom filters. As counter-measures, hardening techniques have been developed that modify the bit patterns in Bloom filters. One recently proposed hardening technique is BLoom-and-flIP (BLIP), which randomly flips bit values according to a differential privacy mechanism. However, while making Bloom filters more resilient to attacks, applying BLIP can lower linkage quality. We propose and evaluate a reference values based BLIP mechanism which ensures that Bloom filters for similar encoded sensitive values are modified in a similar way, resulting in improved linkage quality compared to standard BLIP hardening.