Secondary blocking method in privacy-preserving record links

Abstract

The invention discloses a secondary partitioning method in privacy protection record link, The invention belongs to the field of data integration and data privacy, in particular to the Bloom Filter coding of the records by each data source. Then, the following two steps are carried out: (1) LSH combined with suffix quadratic partitioning method, and block dispersion degree is introduced to adjusttwice partitioning. (2) Multi-block merging based on sliding window can improve the fault tolerance of links. The PPRL block dividing method of the invention has the characteristics of high recall rate of the LSH method and fast division of a large data set, and simultaneously effectively improves the precision rate.

Description

technical field [0001] The invention belongs to the field of data integration and data privacy, and mainly relates to a secondary block method applied in privacy protection record linking. Background technique [0002] With the advent of the big data era, there is an increasing need to mine valuable information by analyzing datasets containing millions of records, and analyzing large datasets usually requires integrating data from multiple sources. At the same time, many organizations do not allow other organizations to share their datasets due to regulations and laws. To this end, the 'Privacy-Preserving Record Linkage (PPRL)' technology is presented, which refers to identifying records representing the same entity in different datasets without revealing the privacy of the entity. In the PPRL project, if a record of a data party is judged to match the records of other data parties, the data party agrees to inform other participants or additional researchers of some attribu...

AI Technical Summary

Problems solved by technology

[0004] There are two deficiencies in the existing PPRL technology at present: 1) Most of the existing PPRL technologies are suitable for the record link between two data parties, but there are few researches on PPRL technology between multiple data parties. The multi-party PPRL method must ensure that the data party The increase in the number will not seriously affect the recall and precision

2) None of the existing block methods can make PPRL achieve high recall rate and high precision rate at the same time, mainly due to the following two aspects: On the one hand, too many unmatched real cases will be generated after block Candidate record groups, resulting in additional calculation costs; on the other hand, the real matching record groups are lost after block, and matching calculations are not performed on them

Images