关键词: 记录链接, 隐私保护, 二次分块, 记录匹配

Abstract: Linking records that represent the same entity among different databases while protecting the privacy of entities stored in those databases is one of the core technologies for the safe and efficient integration of multi-party data resources. However, the existing blocking methods for privacy-preserving record linkage (PPRL) cannot ensure the recall and the precision at the same time, the strong privacy matching methods cost a lot of time and there are few matching studies for more than two databases. To address these problems, this paper proposes a multi-party strong-privacy-preserving record linkage method (MP-SPPRL). Firstly, a double blocking method with locality sensitive Hashing (LSH) blocking and suffix blocking is proposed, which uses dispersion to adjust the double blocking process. The method effectively improves the recall and precision of MP-SPPRL. Next, a sliding window is used to merge the blocks to generate candidate record groups for ensuring the fault-tolerant rate of MP-SPPRL. Then, a scalable multi-party matching algorithm based on secure multi-party computation (SMC) is designed by using the homomorphic Hamming distance calculation for large data. The algorithm significantly reduces the record matching cost by cutting encrypted records and stopping the distance calculation of candidate record groups that cannot be matched. Finally, the empirical evaluation demonstrates the high recall, high precision and high efficiency of MP-SPPRL.

Key words: record linkage, privacy preserving, double blocking, record matching