Table with multilevel generalisation and the use of 4 anonymous, the mention of having greater than or equal to 4 elements per class holds true here

In addition to generating a hierarchy for the quasi-identifiers, the sensitive attributes also have a hierarchy associated with them

Identity disclosure : Where the identity of an individual is obtained from the released table

Attribute disclosure : When the senstitve attribute value is disclosed for an entity, due to problems like homogeneity in the sensitive value attributes in the equivalence class

Along with the 4 unique points mentioned, is adjusting the value of k across the releases of the data also a good idea?

Currently, the best working algorithm is, based on local recoding and fixing the size of the equivalence classes to be multiples of k

SOME IDEAS

What is an inference table?

The most specific values from all releases when written gives the inference table, this can reveal more knowledge than what is needed, and can lead to attacks,