Linkage on Private Data with Secure Multiparty Computation

252 / BBB

2025-09-23

M. S.

15:45

Privacy-preserving Record Linkage is the task to identify and merge records that belong to the same real-world entity within multiple different databases, without revealing any identification details of this real-world entity. This technique can help to gather a lot of data from different sources without violating laws like the General Data Protection Regulation. Use-cases are for example accumulation of meaningful data for medical studies or identification of cash flows to detect financial fraud. Privacy-preserving Record Linkage with more than two parties is carried out by a trusted third party in the current literature. This makes the trusted third party an attractive target for attacks and there might be situations where a trusted third party can not be agreed upon. To combat this issue, we give three protocols to transform already established 2-party Record Linkage protocols to n-party Record Linkage protocols in an honest-but-curious setting. As most protocols in the field of medical Record Linkage use Bloom filters as pseudonymisa- tion data structure, we additionally investigate different Bloom filter generation techniques and introduce another type of Bloom filter, based on already well-established generation variants. Our approach increases the achievable Linkage Quality consistently and can be used in any Record Linkage protocol based on Bloom filters.