An de-identification technique that replaces a value with another pre-defined value. The mapping of original values to replacement values is deterministic. That is, it is defined upfront, and thus allows for a high degree of control over the output data. However, this also means the mapping itself should be protected and not be disclosed externally.

Substitution can also provide a form of generalization, where the substituted value is less precise than the original and several values map to it.

For example, the value iPhone X could be mapped to be substituted by value iPhone – Type 1, and iPhone 8 could be substituted by iPhone – Type 2.

An alternative substitution method could be to map iPhone X and iPhone 8 to the value iPhone.

Return to glossary