How it works
PrivHaven scrubs PII out of CSV exports and hands you evidence of what it did. This page covers what happens to your file, what the tool can and cannot see, and how someone else can check the result.
Your file never leaves this device
The detection engine is WebAssembly running in your browser tab. When you pick a file, it is streamed through that engine and written back out through your browser's own save dialog. It is never held whole in memory and never sent anywhere.
You do not have to take that on faith. Open your browser's developer tools, switch to the Network tab, and scrub a file: you will see no request carrying its contents. The only network calls PrivHaven makes at all are to its own origin, for licence checks.
Three steps, zero uploads
- Scan. Column-aware detection reads the file and reports what it found, including the columns it is unsure about. Scanning is unlimited on every tier, including Free, so nothing is hidden behind payment at this stage.
- Decide. You pick a treatment per column rather than per file: redact, mask, tokenize, drop, or flag it for a human to look at. Nothing is changed until you choose.
- Scrub and export. The cleaned CSV downloads locally. On Pro, so does the compliance report and the tokenization key, which is generated on your device and never transmitted.
What it detects
Seven types, validated rather than merely pattern-matched:
- US social security numbers
- Credit card numbers
- Email addresses
- Phone numbers
- IBANs
- ABA routing numbers
- Dates of birth
Where a format carries a checksum, PrivHaven verifies it. A sixteen-digit number that fails the card checksum is not reported as a card, which keeps the findings list short enough to actually review.
What it does not do
This matters more than the feature list, because a scrubber you trust too much is worse than none.
- CSV only. Spreadsheets, PDFs and JSON are not supported. Export to CSV first.
- No names, addresses or record numbers. There is no named-entity recognition, so a Name or Patient ID column passes through untouched unless you drop or flag it yourself. The report states this rather than implying full coverage.
- No compliance verdict. PrivHaven reports facts about a file. Whether those facts satisfy an obligation is a judgement for your DPO, auditor or counsel.
The report, and why it is checkable
Pro exports a compliance report describing the ruleset that ran, the findings, and the actions taken, bound together in a hash chain. Two properties make it useful as evidence:
- Reproducible. The same file and the same ruleset always produce the same findings, so anyone holding the original can re-derive them.
- Tamper-evident. Editing the report after the fact breaks the chain, and the check that catches it does not require trusting us.
Verification is deliberately independent of this app. The in-browser verifier re-computes everything from the report and your original file, and the same checks are published as a standalone command line tool that contains none of PrivHaven's engine. If our verifier and the independent one ever disagreed, that disagreement would itself be visible.
Evidence, not certification. We are describing an instrument and its readings, not issuing a judgement.
Free and Pro
Free covers unlimited scanning and scrubbing up to 10,000 rows, with redaction, masking and dropping. Pro removes the row ceiling and adds consistent tokenization and the compliance report. Those limits are enforced inside the WebAssembly engine, not in the page, so they are not something a browser tweak can lift.