In the past decades, an important issue of data releasing for further utilization is data privacy. One of the important privacy preservation models,
k-anonymity, is proposed in [
9]. The released data must have at least
k identical tuples for guaranteeing data privacy preservation. In the
k-anonymity process, the link-able attributes [
10], so-called the quasi-identifiers, are to be replaced with the more general data [
13]. The distortion, which causes loss of information, occurs after the data is generalized [
12]. Thus, to ensure both data privacy and minimize information loss, the optimal
k-anonymity is highly desired.
There are a few optimal
k-anonymity algorithms, which have been proposed for preserving the privacy of the generalized dataset [
1,
3,
4]. The existing algorithms are based on the searching through the generalization lattice [
1], the lattice that represents all generalization schemes in the
k-anonymity method, for determining the optimal
k-anonymity solution. The generalization schemes which satisfy the
k-anonymity and loss minimal information will consider the optimal solution. The Flash algorithm [
3] determines the optimal
k-anonymity solutions by binary search on the lattice of generalization. The algorithm searches through all paths in the generalization lattice and terminates when all paths are traversed completely. The incognito algorithm is proposed, in [
4], by dividing the lattice into sub-lattices and searching through each sub-lattice with breadth-first search manners until the search on all sub-lattices is complete. The Optimal Lattice Anonymization (OLA) algorithm is also proposed to address the optimal
k-anonymity problem. The algorithm divides the generalization lattice into sub-lattices and determines whether all sub-lattices satisfy the
k-anonymity condition until the optimal solution is found. These existing algorithms are proposed for the general dataset, a dataset with various datatype. Therefore, in [
6], the authors proposed that there is a special type of datasets that all quasi-identifiers are in the same domain, so-called an Identical Generalization Hierarchy (
IGH) data. With this type of datasets, the special characteristics of the optimal
k-anonymity solution on
IGH data are discovered. The optimal solution of an IGH dataset is always at the lowest level found
k-anonymity satisfied node. Thus, the algorithms for
IGH data privacy-preserving that could utilize this characteristic was proposed in our previous work, Optimal-IGH [
7]. The algorithm is especially proposed for an IGH dataset by performing the pre-order and post-order depth-first search manners. The algorithm can effectively find the optimal
k-anonymity solution on an IGH dataset. However, due to the characteristic of the generalization lattice, the generalization schemes which are considered to satisfy the
k-anonymity condition, all its children are also considered the
k-anonymity condition without visiting it. Thus, the pre-order and post-order manners might not be the most efficient solution for finding the optimal solution.