Before You Share: The 5-Minute Document Check

← PII Shield · All guides

Most data you are asked to share is not secret because it is valuable — it is secret because it identifies someone. A spreadsheet of support tickets, a screenshot of an admin panel, a CV. Here is the short list worth exhausting before you hit send.

Direct identifiers

Anything that names a person on its own: full name alongside contact details, identification numbers issued by a state (NRIC/FIN in Singapore, SSN in the US, Aadhaar in India, BSN in the Netherlands, CPR in Denmark, NINO in the UK), and government business registries like a company UEN. A name without context is often fine; a name with a phone number is a target.

Contact routes

Email addresses and phone numbers. Two same-looking emails are usually the same person, so a masking tool must replace both with the same token — otherwise the reader can still link the two rows to one individual.

Financial keys

Card numbers (check the Luhn digit — a tool should, and ours does), international bank details, and anything your payroll or invoicing flows touch.

Machine secrets

The most commonly leaked class, because it does not look like "personal" data: API keys, bearer tokens, JWTs, connection strings with passwords, and private key blocks. PII Shield treats these as PII — a leaked key identifies and impersonates you. Its detectors are prefix-anchored and checksum-validated, and deliberately quiet on documentation placeholders like sk-your-key-here and example: shouting at a tutorial would train you to ignore warnings.

What checksums buy you

A 16-digit string is a card number only if its Luhn digit checks out. Without that check, masking "detects" card numbers in phone numbers and order ids, and a user learns the tool lies. With it, the finding that remains is the one worth acting on.

A worked example

Take one row of a support ticket: Name: Ada Lovelace · NRIC: S1234567D · Phone: 9123 4567 · Email: ada@example.com. Run it through the checker: the NRIC passes its checksum (so it is real), the phone matches the Singapore mobile shape, and the email is an obvious contact route. Mask all four and the row still reads like a ticket — <NAME_1> · <ID_1> · <PHONE_1> · <EMAIL_1> — but nothing in it identifies anyone, and the token consistency keeps the row joinable to its neighbours. That is the whole trade: you lose the identity, you keep the analysis.

What a leak actually costs

Under Singapore's PDPA and comparable regimes (GDPR in Europe), a leaked spreadsheet of identifiers is a reportable breach with penalties scaled to negligence — not to intent. The practical costs land before any fine: affected people must be notified, the incident investigated, and trust repaired. A five-minute masking pass before sending is cheaper than every one of those steps, and the verify loop is what makes the pass trustworthy rather than assumed.

Then verify the output

Masking once is only half the loop. Re-scan the result. If a token like <EMAIL_1> leaks, the mask used the wrong span. PII Shield's summary reports whether the masked output re-scans clean — that is what "verified" means in this tool.

Everything on this page runs in your browser — open PII Shield and paste the document you are about to send.