07-03-2012, 05:06 PM
(07-03-2012, 11:06 AM)atom Wrote: Hmm a markov-model for whole words? Sounds interessting but would require a bit a different concept. I am not sure yet how this could be done but I think with enough effort its possible, even on GPU. Maybe in a later version and thanks for the input!
Absolutely something different, yes. Then again; my Swiftkey software on my Android phone analyses my sms inbox and more, and then does both word prediction and 'next word' prediction based on the statistics it generates.
Unless you do correct horse battery staples passphrases, but something more logical - even as simple as standard sentences with correct grammar, this would eventually be very effective against natural language passphrases.
I'm not going to claim any original idea here, just going back to the xkcd 936 discussions, including this interesting blog post from @davienthemoose: http://secmoose.blogspot.no/2011/08/why-...-suck.html