sisodiabhumca When to invoke
- When reviewing CSV exports before sharing externally.
- When building a data-loss-prevention (DLP) checklist for analytics datasets.
Inputs needed
--input: Path to a CSV file.- Optional
--out: Path to write a JSON report. - Optional
--max-rows: Max rows to sample (default 2000).
Workflow
- Read CSV headers and sample up to N rows.
- For each column, score risk using:
- Header keywords (email, phone, ssn, address, dob, name).
- Value patterns (email-like, phone-like, IP address, credit card-like, SSN-like).
- Output suspected PII columns with examples (redacted) and suggested remediation.
Output format
JSON with:
suspected_columns: list of{column, risk, reasons, examples_redacted}.summary: counts by risk.
Guardrails
- Redact example values in output.
- Use heuristics only; results are probabilistic.
Reference code
csv_pii_scanner.py