How to redact a PDF so the words are really gone

Why a black box is not redaction

A PDF page is a list of drawing instructions. The words are drawn as characters in a font, and a black rectangle added later is one more instruction, painted over them. A viewer shows the rectangle. Selecting, copying, searching and every program that extracts text read the characters underneath, which are still there.

The same happens one step earlier. Words blacked out in Word with a shape or black highlighting, then saved as PDF, travel into the PDF under the box. Court filings and official reports have been published like this more than once, and readers recovered the text by copying it into another document.

Real redaction removes the characters from the page’s content, and the pixels from any picture under the mark, so there is nothing left to copy.

The same words are often somewhere else too

Taking the words off the page is half the job. A PDF keeps text in places that never appear on a printed page:

• Properties and metadata. The title, subject, author and keywords, and an XMP copy of them that many programs write as well.
• Bookmarks. The outline down the side often repeats headings, which is where names and case numbers sit.
• Comments and their replies, form field values, link addresses, attached files and page labels.
• Earlier versions of the file. A PDF can be saved by adding the changes at its end, which keeps the old bytes inside. A page edited that way can still hold its original content.
• Scans. On a scanned page the words are pixels, so the picture itself has to change. If text recognition was run on it, there is an invisible text layer as well, and that has to go too.

How to redact properly

Work on a copy, and keep the original somewhere safe. Then find every instance before you mark anything: search for each name, number and spelling, and look in headers, footers, tables and captions, where a search is easy to forget.

• Adobe Acrobat Pro has a Redact tool that marks words or areas and applies them, and offers to remove hidden information from the file in the same step. The free Acrobat Reader does not redact.
• LibreOffice redacts a document you have the original of (Tools, Redact) and exports a PDF in which every page is a picture. No text survives anywhere, though the pages can no longer be searched or copied, and the file is larger.
• Turning every page into a picture is the blunt version of the same method. It removes the text layer, but the properties, bookmarks and attachments come through unless you strip them as well.

Online redaction sites work, but they mean uploading the very words you want removed to somebody else’s server. For anything confidential, use a program that runs on your own computer.

Save the result as a new file rather than over the old one, so no earlier version is carried inside it.

Check the result before you send it

Whatever did the redacting, check the file it produced, ideally in a different PDF reader from the one that made it:

• Select everything on each page, copy it and paste it into a plain text editor. Nothing that was redacted should appear.
• Search the file for every word and number you removed.
• Open the document’s properties, the bookmarks, the comments and the attachments, and read them.
• On scanned pages, try selecting the area under each mark. If words highlight, a text layer is still there.

Searching the file’s bytes in a text editor is not enough by itself, because most of a PDF is compressed and the words would not show even if they were present. The checks above read the file as a reader does, which is what the person you send it to will do.