Privacy preserving linkage using multiple match-keys.

Area of Science:

• Data linkage
• Privacy-preserving techniques
• Database security

Background:

• Existing privacy-preserving linkage methods have limitations: anonymous linkage codes offer limited accuracy, while Bloom filters are vulnerable to frequency-based attacks.
• There is a need for robust methods that balance data accuracy with privacy protection.

Purpose of the Study:

• To present and evaluate a novel privacy-preserving record linkage protocol.
• To combine the accuracy of Bloom filters with the privacy benefits of anonymous linkage codes.

Main Methods:

• Developed a protocol using multiple match-keys per record, with key composition tailored to dataset attributes.
• Evaluated the protocol via de-duplication of administrative and synthetic datasets, comparing results against un-encoded linkage and other privacy-preserving techniques.

Main Results:

• The multiple match-key protocol demonstrated high linkage quality across diverse datasets.
• Performance surpassed record-level Bloom filters and the standard linkage key (SLK), though it was outperformed by field-level Bloom filters.

Conclusions:

• The novel protocol achieves high linkage quality while mitigating frequency-based attacks inherent in Bloom filter methods.
• This approach shows significant promise for practical applications in real-world data linkage scenarios.