How do I know my redaction actually worked?
This is the question nobody answers properly. There are hundreds of guides explaining how to black something out and almost none explaining how to find out whether it worked, which is strange, because the second question is the one that decides whether you have a problem.
The honest answer is that you cannot tell by looking, and that the check most people are given only tests one of the ways it goes wrong.
Why looking at it tells you nothing
A PDF is a container rather than a picture. What appears on your screen is one of the things inside it, assembled for display. The words you blacked out can still be sitting in that container, underneath the rectangle, and the rectangle is drawn on top exactly as a sticker would be.
Every program most of us use has taught us the opposite. A commenter on a legal blog put it better than any security guide has managed:
Users are trained WYSIWYG. What you see is what you get. Quoted by Eugene Volokh, writing about redaction failures in court filings
So the page looks finished, and people send it. The failures that reach the news are almost never carelessness. They are people checking the only thing they knew to check.
Three checks, from weakest to strongest
1. Look at the page
Worth thirty seconds, because it catches the mistake nobody expects to make: a box in the wrong place, a second mention of the name further down, an account number repeated in a footer. Read the whole document rather than the part you were working on. If it is a bank statement, there is a page on exactly that job.
What it cannot tell you is anything at all about what is in the file, because the file is not what you are looking at.
2. Select all, copy, paste into a plain text editor
Better than dragging across the black box, because it pulls the whole text layer out rather than whatever happens to sit under your cursor. If the name you removed appears in the pasted text, you have your answer immediately.
A clean result is where people stop, and it is where the trouble starts. This test does not see a hidden text layer behind a scan, an earlier version saved inside the file, the document properties, the bookmarks, the page thumbnails, an attachment, or a form field still holding its value. We wrote those out in full, with what each one has actually leaked, in the copy-paste test is not enough.
3. Inspect the file itself
Everything above is inspectable. Most people have no practical way to do it, which is why we built a free tool that does it in one pass. Drop a PDF in and it reports on each of those hiding places by name, and where it finds text it should not, it shows you the text on screen. It also writes a short report you can keep, which records what was found and where without reproducing any of it — a report that quoted your document back at you would be a second copy of the thing you were protecting.
It reads the file and nothing else. It never writes a new one, it never produces a redacted version, and your document stays on your machine — you can switch your wifi off and it works exactly the same.
What none of this can check
Two things, and both are yours to judge.
Whether you covered the right things. No tool knows which words were the sensitive ones. A document can pass every check in existence and still have the name in the second paragraph that you meant to remove and forgot.
Whether the picture itself gives it away. Once a page has been rebuilt as an image, there is no text left to recover, and the checks come back with nothing to report. That is the right answer for text, and it says nothing about what the image shows. A box that stops one letter short is still legible, and only your eyes will catch it.
If the check comes back bad
- Do not repair the file you have. Drawing a second box over the first leaves both, and adding to a file is how earlier versions end up travelling inside it. Go back to the original.
- Redo the redaction with something that removes rather than covers. The words have to leave the file. Anything that draws on top of them has not done the job, whatever it is called in the menu.
- Check the new file, not the method. Software that redacted one document correctly can fail on the next one, which is the point of checking output rather than trusting a brand.
Before you send it
Check the document — free, no account, nothing uploaded. If it finds something, it shows you what and where.
This is not legal advice and it certifies nothing. What it tells you is what is in your file and what your last tool did or did not take out of it.