A deterministic approach for protecting privacy in sensitive personal data.

Area of Science:

• Data privacy and security
• Statistical methodology
• Biomedical informatics

Background:

• Data privacy is a major challenge in processing personal and sensitive medical research data.
• Sharing sensitive data is problematic, conflicting with the open data principle vital for scientific advancement.
• Existing statistical methods and computational tools aim to aid data custodians in overcoming privacy challenges.

Purpose of the Study:

• To propose a novel deterministic approach for anonymising personal data.
• To address the conflict between data privacy requirements and the need for data sharing in research.

Main Methods:

• A deterministic anonymisation method is proposed.
• The approach stratifies data by categorical variables.
• Continuous variables are re-distributed using a k-nearest neighbours algorithm.

Main Results:

• The method was demonstrated on real-world datasets, including Titanic passenger data and the 1958 Birth Cohort.
• The anonymisation procedure effectively hinders data re-identification.

Conclusions:

• The proposed method makes data re-identification difficult.
• It minimises the loss of data utility by preserving spatial properties.
• Informative statistical analysis can still be conducted on the anonymised data.