PROJECT TITLE :

FEDERAL: A Framework for Distance-Aware Privacy-Preserving Record Linkage - 2018

ABSTRACT:

In privacy-preserving record linkage, a number of knowledge custodians encode their records and submit them to a trusted third-party who is accountable for identifying those records that confer with the same real-world entity. During this Project, we tend to propose FEDERAL, a novel record linkage framework that implements ways for anonymizing each string and numerical data values, which are usually gift in knowledge records. These strategies depend on a robust theoretical foundation for rigorously specifying the dimensionality of the anonymization area, into which the initial values are embedded, to produce accuracy and privacy guarantees underneath various models of privacy attacks. A key part of the applied embedding process is the edge that is needed by the space computations, which we prove can be formally specified to ensure accurate results. We evaluate our framework using three real-world knowledge sets with varying characteristics. Our experimental findings show that FEDERAL offers a whole and effective resolution for accurately identifying matching anonymized record pairs (with recall rates constantly higher than ninety three %) in massive-scale privacy-preserving record linkage tasks.