Anatomisation with slicing: a new privacy preservation approach for multiple sensitive attributes.

Area of Science:

• Health Informatics
• Data Privacy
• Cybersecurity

Background:

• Vast amounts of personal health information (PHI) are generated, posing significant risks if compromised.
• Existing anonymization methods (e.g., generalization, bucketization) are inadequate for high-dimensional data with multiple sensitive attributes (SA).
• Privacy threats like membership, identity, and attribute disclosure are major concerns in healthcare data management.

Purpose of the Study:

• To propose a novel anonymization technique for high-dimensional datasets with multiple sensitive attributes.
• To enhance data privacy by adhering to k-anonymity and l-diversity principles.
• To minimize information loss and computational complexity while preserving data utility.

Main Methods:

• A hybrid approach combining anatomization and an enhanced slicing algorithm.
• Anatomization dissociates quasi-identifiers from sensitive attributes (SA), creating separate tables.
• Enhanced slicing uses vertical partitioning, clustering, and tuple partitioning (MFA) to reduce dimensionality and group correlated SAs.

Main Results:

• The proposed method successfully preserves privacy for data with numerous sensitive attributes.
• Anatomization minimizes information loss, while the slicing algorithm maintains data correlation and utility.
• Advanced clustering algorithms significantly reduce processing time and complexity.

Conclusions:

• The integrated approach effectively addresses the limitations of existing methods for high-dimensional, multi-sensitive attribute data.
• The technique provides robust protection against privacy threats, including membership, identity, and attribute disclosure.
• This method offers a practical solution for secure management and analysis of sensitive health information.