I seem to recall an anecdote on how one branch of the British government had designed a way to circulate seemingly identical for-your-eyes-only documents to a small circle of high-level collaborators, while ensuring the possibility to identify them if the document leaked. The story was that they used tiny variations in typography to make differentiable documents that seemed identical to the naked eye.
So, my question is: Is that feasible? Does that involve modification of font outlines, or kerning, or something else? Is that done in the font itself, or in the typographic engine?
Yes, there are methods to do this. It’s not used with the typography though (speaking in general). It’s related to steganography and the common method is to use yellow ink to print a pattern of dots which can be recognized as a serial number or other ID depending on purpose.
It is printed in a small size typically close to an edge. As yellow is hard to read for the human eye on white paper, this is rarely discovered. To detect it blue light is used together with scanning.
This is not only used by government, but is common in consumer printers from manufacturers such as Brother, Canon, Dell, Epson, HP, IBM, Konica Minolta, Kyocera, Lanier, Lexmark, Ricoh, Toshiba and Xerox (2).
I choose to put my comment here as part of the answer even though if will be of speculative nature as the actual facts are not available to the public.
As for other techniques related to intelligence/security services I would assume that this information would still be classified (details of the techniques). We can only speculate.
Too small variations (ie. small “artifacts” in the glyphs) would be lost due to quality loss in the copy process, so it need to be related to geometry and the relationship between the glyphs, ie. spacing, difference in heights, line-heights, use of different words at different locations etc. Perhaps a checksum/acid-test mechanism to handle small errors. I’m just trowing out possible scenarios here.
If it is to a small group you wouldn’t need more than a few chars to encode this sort of information (ID). And it’s not so difficult to write a software that implement this.
Source : Link , Question Author : F’x , Answer Author : Community