Why a black rectangle is not a redaction
A PDF is not a picture of a page. It is a content stream of drawing
instructions, and the text sits in that stream as text. When a markup
tool draws a filled rectangle over a name, it appends one more drawing
instruction on top. The name is still in the file, in order, fully
selectable. Copy and paste from the "redacted" area, run
pdftotext over it, or open the file in almost any editor,
and it comes straight back out.
This is the single most common way documents leak. It has produced published court filings with witness names recoverable by selecting the text, and public-records releases where the redactions came off with a drag of the mouse. The document looked right on screen every time.
The same trap catches the flatten-to-image workaround. Rasterising a page does destroy the text layer, but it also destroys every other thing that made the file useful: selectable text, accessibility, search, file size. And it leaves the document metadata untouched.
What DokMine does instead
DokMine applies true redaction. Where a match is found, the text is removed from the content stream and the region is covered — the glyphs are not painted over, they are taken out. The result is a PDF, not an image of one: the pages that were not touched keep their selectable text, their structure and their fonts.
You choose how the removed values are represented. Blackout
leaves a solid bar. Tokens substitute a typed
placeholder — [EMAIL_1], [PHONE_2] — with the
same value mapping to the same token throughout the document, so a
redacted contract still reads as a contract and you can still tell that
two mentions referred to the same party.
The places PDFs hide personal data
- Scanned pages and embedded screenshots. Text inside an image is invisible to a text search. DokMine runs OCR over embedded images and blacks out matches in the image data itself.
- Document metadata. Author, title, subject, keywords and producer fields routinely carry a real name long after the body has been cleaned.
- Annotations and form fields. Comments, sticky notes and filled form values live outside the page content stream and survive a body-only pass.
- Headers, footers and stamps. Repeated on every page, easy to skim past when reviewing.
How to redact a PDF with DokMine
- Drop the file in. One PDF or a batch. No install, no plugin, no Acrobat licence.
- Pick what to look for. Emails, phone numbers, URLs, person names, card numbers, IBANs, IP addresses — enable only the types that matter for this document.
- Run extraction first if you want to look before you leap. Extraction mode lists every match, per page and per type, so you can confirm the selection is right.
- Redact and download. Choose blackout or tokens, then open the returned PDF and read it before it goes anywhere.
Check the result — every time
No automated detector is exhaustive, and PDFs are the format most likely to defeat one: skewed scans, unusual fonts, text split across drawing operations, handwriting, tables reconstructed from absolute positions. Names in particular are ambiguous by nature and will never be caught reliably by any tool, DokMine included.
Treat the output as a strong first pass. Open the redacted PDF, read it, and try selecting text over the redacted regions. If DokMine could not read an embedded image, it says so rather than handing the file back as though it had been checked.