BLOOM FILTER BASED SECURE RECORD LINKAGE: CRYPTANALYSIS ATTACK METHODS

In the field of data science, integrating and analyzing data distributed across different data sources is necessary for decision-making and planning business strategies. Hence, the record linkage is important component in data integration and analytics for matching and linking records across various databases. Since databases contain personally identifying information and sensitive data about individuals, there is a need for protecting privacy for record linkage. Thus, the secure record linkage involves encoding of records from multiple data sources and then performs matching on them. The Bloom filter encoding techniques were found to provide better privacy while performing approximate matching on erroneous records. Still the Bloom filter based secure record linkage techniques can suffer from re-identification attacks and may lead to imbalance between linkage accuracy and privacy. This work includes the overview of secure record linkage and implementation of recent attack methods on EPPRL approach, Basic and Balanced Bloom filter. Moreover, we provide recommendations to limit the vulnerability and re-identification of encoded identifiers from Bloom filter based secure record linkage.