Talks about the partitioning of the dataset into k-anonymous groups, in nlogn time

Single dimension means, the partition so formed must result in k-anonymity across all partitions

Multidimensional requires that only that partition can be divided so as to ensure k-anonymity

The idea is at every step try to divide the region by 2 or at the median of that region and see if there are k values or not, check in each dimension, till it is possible

When minimal cuts are achieved for each attribute, it will be stopped for that dimension, minimal cuts will ensure that u have the maximum data quality

Replacing the range with the mean is also an interesting approach, where data quality can be retained, it adds some ambiguity

In every dimension check for a split of that attribute obeying k anonymity across that dimension alone and then repeat till minimal partitioning.

Split into 2 parts not necessarily median, at some optimal point.

checking for k-anonymity at that and l-diversity in the cut made, each attribute is independent of the other.

Maintaining multiples of k.