Conclusion

Clean up !

{

DROP EXTENSION anon CASCADE;

{


REASSIGN OWNED BY jack TO postgres;
REVOKE ALL ON SCHEMA public FROM jack;

REASSIGN OWNED BY paul TO postgres;

REASSIGN OWNED BY pierre TO postgres;

{

DROP DATABASE IF EXISTS boutique;

{

DROP ROLE IF EXISTS jack;
DROP ROLE IF EXISTS paul;
DROP ROLE IF EXISTS pierre;

Many Masking Strategies

Static Masking : perfect for "once-and-for-all" anonymization
Dynamic Masking : useful when one user is untrusted
Anonymous Dumps : can be used in CI/CD workflows
Generalization : good for statistics and data science

Many Masking Functions

Destruction and partial destruction
Adding Noise
Randomization
Faking and Advanced Faking
Pseudonymization
Generic Hashing
Custom masking

RTFM -> Masking Functions

Advantages

Masking rules written in SQL
Masking rules stored in the database schema
No need for an external ETL
Works with all current versions of PostgreSQL
Multiple strategies, multiple functions

Drawbacks

Does not work with other databases (hence the name)
Lack of feedback for huge tables (> 10 TB)

Also...

Other projects you may like

pg_sample : extract a small dataset from a larger PostgreSQL database
PostgreSQL Faker : An advanced faking extension based on the python Faker lib

Help Wanted!

This is a free and open project!

labs.dalibo.com/postgresql_anonymizer

Please send us feedback on how you use it, how it fits your needs (or not), etc.

This is a 4 hour workshop!

Sources are here: gitlab.com/dalibo/postgresql_anonymizer

Download the PDF Handout

Questions?

:::

::: {.cell .code}

python

:::