An Enhanced Approach for Privacy Preserving Record Linkage during Data Integration

Abstract

Today collecting and integrating data from multiple datasets has become a vital part to perform various analysis tasks. Record linkage plays an important component during data integration to detect and link similar data instances. However, personal and sensitive data need to be protected in a manner so that there is no re-identification of original attribute values by the party performing record linkage. Now-a-days, the Bloom filter encoding has utilized across many countries for privacy preserving record linkage. Moreover, the hardened approaches of Bloom filter encoding enhance privacy at the cost of reduced linkage accuracy. Still the security concerns remain with the Bloom filter encoding techniques because attackers can re-identify the obfuscated data with the use of available public resources. We propose an enhanced approach for privacy preserving record linkage (EPPRL) during data integration to achieve better privacy with acceptable linkage accuracy. The results show that the proposed approach EPPRL outperforms in comparison with Balanced Bloom filter encoding technique in terms of precision, recall, f-measure and re-identification of attribute values.