Other than it being color, that's what I would have assumed was done here, except the staple holes and relatively decent image quality do seem to argue against it.
Other than it being color, that's what I would have assumed was done here, except the staple holes and relatively decent image quality do seem to argue against it.
The simplest way to get around that would be to retype the whole document, which is a bunch of friction. But to counter that, more invasively, the texts different people would get could be different themselves. Sentences subtly reordered, slight differences in word choice or spelling. Though that makes collaboration more difficult.
Less technical savvy but resourced organizations wouldn't face difficulty using those approaches. It could just be a bit that an IT admin toggles on the GSuite or MS Office admin panel.
I think that would make collaboration impossible. If that were standard practice collaboration would eventually uncover that was occurring, and a leaker would compensate.
IMHO that technique is only valid when the recipients can be assumed to not be in contact, or as an unusual thing to identify a leaker when one is already suspected.
I don't think that's really the issue. Let's suppose you have 64 employees with access to a doc. You can uniquely identify them with 6 bits in the doc. A doc of a couple pages would be more than enough to put in those bits, and many more. Even if you had two collaborators who shared docs for a word for word diff, they'd still miss half the identifying bits, which reduces the potential culprit pool to 8, and that's assuming zero redundancy.
The issue I imagine is more usability. Someone identifies one of those bits as content that should be changed: what happens when they suggest the edit?
That's not really what I'm talking about. It seems like you're focusing on identifying all the exact bits that are being used to identify someone, but that's not important. What is important is the recipients figuring out that someone is monkeying with the documents in a systematic way to trace them. Once a leaker knows that, instead of leaking the exact document for publication, they'd either leak reworded summary of key points themselves, or require the journalist to only publish something similar.
I think the reason you don't do that is that things like the formatting and so on being accurate lend credibility to the document.
OCR also isn't going to be 100% though probably more than good enough in this case. If you really want to implement subtle tracking, you could make changes in the document itself that differ by recipient.