Oh, this is a nifty hack. Researchers analysing documents with blacked-out words can work out what the words are by measuring the pixel width of the obscured word and working out which letter combinations fit it. A bit of contextual analysis can disregard any words that don’t fit in and boil it down to a few variations, which can then be processed by a human to pick out the most likely candidate. A nice and clever combination of typography and natural language processing, though ultimately there is a very simple security precaution that can prevent this, namely print sensitive information in a fixed-width font.

(via Boing Boing)