A study of improved k-anonymity privacy-preserving algorithms for data sharing

Abstract In response to the limitations of existing clustering-based k-anonymity models in handling outlier data and the insufficient consideration of balancing information loss and data utility, an optimal clustering-based k-anonymity algorithm (OCBKA) was proposed. By transforming the optimization problem of k-anonymity utility into the optimal clustering problem of anonymous equivalence classes, OCBKA employed the Local Outlier Factor algorithm to rank data based on outlier scores, thereby identifying and separating outliers. Simultaneously, by constructing natural equivalence classes, the algorithm adopted a clustering approach to select optimal cluster centers, completing the clustering process. The distances generated during cluster formation are retained to enhance the algorithm’s efficiency. Experimental results demonstrated the algorithm’s effectiveness in reducing information loss and improving execution time, significantly reducing the impact of outlier data on information loss.