Designing a Novel Approach Using a Greedy and Information-Theoretic Clustering-Based Algorithm for Anonymizing...

Area of Science:

• Computer Science
• Data Privacy
• Information Security

Background:

• Data anonymization is crucial for privacy but faces a trade-off between privacy enhancement and data utility.
• K-anonymity is a standard technique, but it is vulnerable to attribute disclosure when sensitive values lack diversity.
• T-closeness offers robust privacy protection against attribute disclosure.

Purpose of the Study:

• To propose a novel clustering-based algorithm for data anonymization.
• To achieve strict privacy protection by satisfying both k-anonymity and t-closeness constraints.
• To balance privacy preservation with data utility by minimizing attribute modifications.

Main Methods:

• A greedy and information-theoretic clustering algorithm is employed.
• Data is clustered based on quasi-identifier similarity and sensitive attribute diversity.
• Clusters are adjusted (split/merged) to meet k-anonymity and t-closeness requirements.
• Quasi-identifier values are replaced with cluster center values for anonymization.

Main Results:

• The proposed algorithm successfully achieves k-anonymity and t-closeness.
• Experimental results show minimized modifications to attribute values, preserving data utility.
• The approach was validated on microdata sets from Facebook, Twitter, and Google+.

Conclusions:

• The novel clustering algorithm effectively enhances data privacy while maintaining data utility.
• The method provides a robust solution for anonymizing sensitive data, addressing limitations of traditional k-anonymity.
• This approach offers a practical method for privacy-preserving data release in social media contexts.