Notes and Warnings

Note

In the case of multiple sensitve attributes, two approaches can be considered:

• Calculate the different properties for each sensitive attribute individually, and take the maximum or minimum as appropriate (for example, the minumum value of \(\ell\) for \(\ell\)-diversity and the maximum value for \(\alpha\) in the case of (\(\alpha\),k)-anonymity).

• Calculate the different properties for each sensitive attribute individually but modifying the set of quasi-identifiers by adding to this set, in addition to the initial quasi-identifiers, all the sensitive attributes except the one under analysis. Then, as in the previous approach, the minimum or maximum of each parameter as appropriate is taken. It is important to note that since the set of quasi-identifiers is updated each time the calculations are made for each SA, the computational cost is much higher in this case.

Specifically, in order to address this challenge, a parameter gen is introduced in all functions except k-anonymity (not applicable). If gen=True (default value), the process of the first approach is followed: generalizing. Otherwise, the second approach is followed, updating the quasi-identifiers.

Warning

It’s important to take into account that the values of t and \(\delta\) for t-closeness and \(\delta\)-disclosure privacy respectively must be strictly greater than the ones obtained using pyCANON (see the definition of that techniques).