Parity: SSN detection misses Unicode hyphens and line-wrapped identifiers

[jzonthemtn]

Description

The formatted SSN patterns in ssn_filter.py handle ASCII hyphens, unformatted numbers, and ASCII spaces. They do not cover nonbreaking hyphens (U+2011), or an ASCII hyphen followed by a newline between groups.

Fixture difference: the Java audit used 078-05-1120, but Python explicitly excludes that identifier. Use 123-45-6789 for this parity regression so the formatting fix does not require weakening the existing exclusion rules. The exclusion itself is not the defect reported here.

Expected behavior and parity

Track parity with philterd/phileas#385, opened for finding R4 from the Philter 4.0 release audit. Recognize supported Unicode separators and line wrapping within an otherwise valid SSN. Redaction must cover the complete original identifier and preserve surrounding text.

Use these fabricated test inputs (decode the escapes before filtering):

Case	Input
ASCII control	SSN: 123-45-6789
Nonbreaking hyphens	SSN: 123\u201145\u20116789
Line wrapping	SSN: 123-45-\n6789

The two variant gaps were identified by source inspection in this implementation; an end-to-end runtime reproduction has not yet been performed. Confirm the behavior with regression tests before implementing the correction.

Acceptance criteria

• Confirm the ASCII control and reproduce both variant failures through the detector and redaction pipeline.
• Align supported Unicode separators and line-wrap patterns with the behavior agreed in SSN detection misses Unicode hyphens and line-wrapped identifiers phileas#385.
• Preserve accurate offsets into the original text if normalization is introduced.
• Add negative cases to prevent matching unrelated numbers across lines and preserve existing SSN validity/exclusion rules.
• Verify detection spans and final redacted output, including repeated identifiers and surrounding Unicode text.
• Document intentional exclusions.