Hardening Encrypted Patient Names Against Cryptographic Attacks Using Cellular Automata

Abstract

Linking information across different databases enables new research in the medical sciences. Recent EU privacy regulations recommend encrypting personal identifiers used for linking. In this contribution, a new method for hardening such a privacy-preserving record linkage technique (PPRL) against attacks is presented. The new hardening method prevents re-identifications and cryptographic attacks while still delivering acceptable linkage quality. Using real-world mortality data, we compare clear-text and several current PPRL methods with our newly proposed method. While all PPRL methods will have to balance security and quality, the use of a cellular automata transformation to protect against attacks will decrease the linkage quality only slightly, while preventing all currently known methods of decrypting Bloom filter-based private linkage keys.