anonymize rdbms datas, keeping key relations

Question

Is there a "ready-to-use" method to anonymize datas, but keeping relations between keys ? For example, I have :

• Table #1

user code	zip code
ztxp15	45789

And :

• Table #2

user code	order date
ztxp15	2021-06-27 06:22pm

I want it anonymized as :

user code	zip code
xvdf65	32165

And :

• Table #2

user code	order date
xvdf65	2021-06-27 06:22pm

This would need : a bijective function that transform a data, keeping its format ([a-z]{4}[0-9]{2}), generating the same value, according a passphrase for example. In this way, unicity will be kept, format too, etc. But maybe I miss something. I think that this problematic is very common so I am looking for previous work about it.

Answer 1

It is a common practice to use a user identifier, which my itself has no meaning to a viewer. I assume in your case this is the user code.

You should only anonymise PII (Personally Identifiable Information). You can encrypt it for bi-directionality, or hash it for single direction anonymise. Hashing is usually done when exporting data to analytics dashboards.

It is not a common practice to anonymise user code. If all PII is anonymised, then the user code is effectively anonymised.