TopDown-KACA: An efficient local-recoding algorithm for k-anonymity

K-anonymity is an effective model for protecting privacy while publishing data. KACA algorithm is a typical generalization algorithm for k-anonymity, which can generate small information loss, but its efficiency is low, especially when dataset is large. Another generalization algorithm, topDown, has high efficiency but generates heavy information loss. In this paper, we propose an efficient generalization algorithm for k-anonymity, called topDown-KACA, which combines the topDown algorithm with the KACA algorithm. The idea of topDown-KACA algorithm is to partition the whole dataset into some subsets by topDown algorithm at first, and then k-anonymize these subsets by KACA algorithm respectively. Experiments show that the proposed algorithm is more efficient than KACA algorithm with similar information loss, and generates less information loss than topDown algorithm with similar execution time.