Hidden in “plain text” when <i>rendered using their special font</i>. The special font slightly changes the shape of each letter to encode additional information. Title is pretty clickbait. If you copy the “plain text” to ASCII or use a regular font this won’t work.
This could be a useful variation on a trick for tracking
leakage of documents: distribute N paper "copies" of a
document to a pool of suspected leakers, each "copy" with tracking data embedded, and when leaks occur, match the
tracking data to the leaker. Old way: use apparent
typos to track, this way: use font variations, next way:
use variations in kerning between words.
Reminds me of an old friend's paper from the late 90's, where you can change the actual words to synonyms to encode messages (rather than the font):<p><a href="http://web.mit.edu/keithw/tlex/lsteg.pdf" rel="nofollow">http://web.mit.edu/keithw/tlex/lsteg.pdf</a>
textmark.io is doing something along these lines but in a different way. It’s a very cool space to be in esp with advancements in machine learning. Disclaimer I am the founder.