Skip to content
Outrich Data

Glossary

l-diversity

l-diversity is a privacy measure that extends k-anonymity. It requires each group of records sharing the same quasi-identifier values to contain at least l different values of a sensitive attribute. This prevents an observer from learning a person's sensitive detail merely by knowing which group they belong to.

k-anonymity protects against singling out a record. It does not protect against learning a sensitive fact. Suppose a group of several records shares the same age band and region, and every record in that group shows the same diagnosis. An observer who knows a person falls in that group learns the diagnosis without identifying which record is theirs. This is sometimes called a homogeneity problem.

l-diversity addresses it by looking inside each group. The sensitive column, such as a diagnosis, a salary band or a disciplinary outcome, must take at least l distinct values within every group. Achieving this may require more generalization or the removal of groups that are too uniform. The measure has limits of its own. It treats all distinct values as equally different, so a group could hold several values that are all similarly sensitive. Further measures exist to refine it. For most company datasets, checking k-anonymity first and then looking for uniform sensitive values in small groups captures the practical point.

The measure only applies where a dataset has a clearly sensitive column. Many operational datasets, such as support threads, carry their sensitive content in free text instead, where review by sampling is the practical control.

Find out if your data qualifies.

Eleven questions, about five minutes, no call required. You get an answer on screen and a written follow-up.

esc