Encoding of Numerical Data for Privacy-Preserving Record Linkage.

Area of Science:

• Computer Science
• Bioinformatics
• Data Science

Background:

• Privacy-preserving record linkage (PPRL) is crucial for sensitive data analysis, especially in healthcare.
• Bloom filters are a popular PPRL technique but traditionally handle string data.
• Existing methods lack efficient encoding for numerical data like geocoordinates.

Purpose of the Study:

• To evaluate a novel numerical encoding method for geocoordinates within Bloom filters.
• To adapt and assess this method for improved privacy-preserving record linkage.
• To enhance the utility of Bloom filters for datasets containing geographical information.

Main Methods:

• Developed a numerical encoding technique specifically for geocoordinate data.
• Integrated this numerical encoding into Bloom filters for PPRL.
• Compared the performance of the proposed method against traditional string-based encoding using synthetic datasets.

Main Results:

• The numerical encoding of geocoordinates achieved superior recall compared to string-based methods.
• The proposed approach demonstrated higher precision in record linkage tasks.
• Demonstrated improved data linkage accuracy and efficiency for geocoordinate data.

Conclusions:

• Numerical encoding offers a significant advancement for Bloom filter-based PPRL.
• This method enhances both the quality and privacy level of record linkage.
• The findings suggest broader applicability of Bloom filters for diverse numerical data types.