How do you generate SSNs without dashes?

[npatki]

Filing this question on behalf of a user from a private thread.

The auto-generated plan sees that a column is PII, but it is SSN without the dashes …. With the sdtype = ssn still work given no dashes or is the a custom generator on our side to be developed?

[npatki]

Assuming that this is the metadata you have for your ssn column:

"my_ssn_column": {
  "sdtype": "ssn",
  "pii": true
}

Then by default, SDV synthesizers will generate random SSN values that contain dashes, for eg. 236-57-5670. This is happening because SDV uses the Faker library for PII anonymization -- and Faker is only capable of producing SSNs without dashes (see the Faker documentation).

The fix: Override the column

Luckily you can override the anonymization method. Instead of using Faker's SSN generator, you supply a generic generator that combines 9 random digits. To do this, create a generic transformer and apply it using the update_transformers function.

from rdt.transformers.pii import AnonymizedFaker

# a generic generator that creates combinations of 9 digits
my_ssn_transformer = AnonymizedFaker(
    provider_name=None,
    function_name='bothify',
    function_kwargs={'text': '#########'}
)

# apply this generator to the ssn column of your synthesizer
synthesizer = GaussianCopulaSynthesizer(metadata)

synthesizer.auto_assign_transformers(data)
synthesizer.update_transformers({
    'my_ssn_column': my_ssn_transformer
})

synthesizer.fit(data)

Now the synthetic data will contain SSNs values without dashes: 375766167