2019/11/28 by Dinusha Vatsalan, Vatsalan, Dinusha, Peter Christen +3
Computer Science · Decision Sciences · #Data Quality and Management #Databases (cs.DB) #Distributed #FOS: Computer and information sciences #Parallel #Privacy-Preserving Technologies in Data #and Cluster Computing (cs.DC)
paper · pdf · doi:10.48550/arxiv.1911.12930
openalex publication_date 2019/11/28 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Privacy-Preserving Record Linkage (PPRL) supports the integration of\nsensitive information from multiple datasets, in particular the\nprivacy-preserving matching of records referring to the same entity. PPRL has\ngained much attention in many application areas, with the most prominent ones\nin the healthcare domain. PPRL techniques tackle this problem by conducting\nlinkage on masked (encoded) values. Employing PPRL on records from multiple\n(more than two) parties/sources (multi-party PPRL, MP-PPRL) is an increasingly\nimportant but challenging problem that so far has not been sufficiently solved.\nExisting MP-PPRL approaches are limited to finding only those entities that are\npresent in all parties thereby missing entities that match only in a subset of\nparties. Furthermore, previous MP-PPRL approaches face substantial scalability\nlimitations due to the need of a large number of comparisons between masked\nrecords. We thus propose and evaluate new MP-PPRL approaches that find matches\nin any subset of parties and still scale to many parties. Our approaches\nmaintain all matches within clusters, where these clusters are incrementally\nextended or refined by considering records from one party after the other. An\nempirical evaluation using multiple real datasets ranging from 3 to 26 parties\neach containing up to 5 million records validates that our protocols are\nefficient, and significantly outperform existing MP-PPRL approaches in terms of\nlinkage quality and scalability.\n