Exploring K-anonymity in genomic datasets: Implications for privacy and survival analysis

This study evaluates K-anonymity methods for protecting sensitive healthcare and genomic data while preserving quasi-identifiers for analysis. It examines techniques such as Nearest Neighbor Clustering-Based Anonymization, Recursive Partitioning with Median Splits, Mondrian K-Anonymity, Value Generalization Hierarchies (VGHs) in Optimal Lattice Anonymization (OLA), Maximum Distance to Average Vector (MDAV), and Top-Down Generalization (TDG). Each method balances privacy, data utility, and computational efficiency. The study compares their effectiveness in minimizing information loss while maintaining privacy using survival analysis tools like the Akaike Information Criterion (AIC) and Concordance Index (C-Index). OLA and TDG provide more generalization, while MDAV and Mondrian are more efficient.