K-Anonymity Privacy Protection Algorithm for Multi-Dimensional Data against Skewness and Similarity Attacks

Area of Science:

• Computer Science
• Data Privacy
• Artificial Intelligence

Background:

• Multi-dimensional data publishing requires robust privacy protection for applications like scientific research and policy-making.
• Existing K-anonymity clustering methods suffer from inconsistent results, low accuracy, and vulnerability to skewness and similarity attacks.

Purpose of the Study:

• To propose a novel K-anonymity privacy protection algorithm (KAPP) for multi-dimensional data.
• To enhance defense against skewness and similarity attacks by combining K-anonymity with t-closeness.
• To improve the accuracy and diversity of data clustering for privacy preservation.

Main Methods:

• Developed a multi-dimensional sensitive data clustering algorithm using an improved African vultures optimization technique.
• Enhanced the optimization by refining initialization, fitness calculation, and solution update strategies for clustering centers.
• Introduced an equivalence class partition method utilizing sensitive data distribution difference values and t-closeness.

Main Results:

• The improved African vultures optimization achieved highly accurate clustering of multi-dimensional datasets across multiple sensitive attributes.
• The KAPP algorithm effectively reduced data similarity within equivalence classes, mitigating skewness and similarity attack risks.
• Experimental results demonstrated KAPP's superior performance in clustering accuracy, diversity, and anonymity compared to existing methods.

Conclusions:

• KAPP offers a robust solution for multi-dimensional data privacy protection against common attacks.
• The integration of improved African vultures optimization and t-closeness significantly enhances data anonymization effectiveness.
• The proposed method ensures data utility while maintaining a high level of privacy in data publishing.