Lessons Learned: Building a Privacy-Preserving Entity Resolution Adaptation of PPJoin using End-to-End Homomorphic Encryption.

Entity resolution is the task of disambiguating records that refer to the same entity in the real world. In this work, we explore adapting one of the most efficient and accurate Jaccard-based entity resolution algorithms - PPJoin, to the private domain via end-to-end homomorphic encryption. Towards this, we present our precise adaptation: HE-PPJoin that details certain subtle data structure modifications and algorithmic additions needed for correctness and privacy. We implement HE-PPJoin by extending the PALISADE (now merged with OpenFHE) open-source, homomorphic encryption library and perform experiments to analyze its accuracy and incurred overhead. Furthermore, we directly compare HE-PPJoin against P4Join, an existing privacy-preserving variant of PPJoin, which uses hashing for raw content obfuscation (encryption), by demonstrating a rigorous analysis of the efficiency, accuracy, and privacy properties achieved by our adaptation as well as a characterization of those same attributes in P4Join. In building and designing HEPPJoin, we faced numerous challenges that required making tradeoffs and analyzing possible alternatives. We have thus summarized and detailed all the lessons we have learned, presented throughout the paper, intended as motivating building blocks for future work in this direction.