Ray Dillinger writes:
Give me a few dozen writing samples from each of a hundred known people, and another writing sample a hundred words long from one of them under a pseudonym, and I can tell you to a 90% probability which of the hundred known people wrote it.
Ray you bring up a very interesting point. Does anyone know of a source for public domain, or open source, writing analysis software that will do just that ? Perhaps using such wares for testing purposes one can craft software to make subtle grammar or spelling or usage alterations in text based communications in order to defeat this sort of analysis. E.G. you've got a remailer or portal of some kind that accepts plain text input, and according to a set of rules makes shifts, sentence length, punctuation usage, spelling usage, and so on, to alter the distinctive signatures associated with the writing patterns of the person in question. THEN afterwards this is piped into an encryption module or just transmitted piecemeal at random timings in order to defeat traffic analysis.