I was going to put this in nuggets, but I figured it had serious enough implications to warrant being in RL issues:

http://www.usatoday.com/tech/news/computersecurity/2005-09-20-keystrokes-passwords_x.htm?csp=34

I had heard about this shit before, but it had a very poor accuracy - nowhere near what these guys are hitting. Anyone out there more educated about this that can verify the validity of their study?

I'd read about that from another source claiming slightly higher accuracy than that even, including all their spelling and grammar algorithms.

You have to think it would be very keyboard dependent though.

I kind of wonder... voice recognition software has gotten to the point where they can do a marginal job without even being trained to a specific person's voice. In fact, I'd surmise that there are less variables involved in keyboard recognition than voice recognition, it'd just be harder to get accurate information out of them.

I wonder though, if it wouldn't depend a lot on how hard you type, the weighting and recoil on your keyboard, and other factors like that; beyond just acoustics.

Yeah. You could account for those things, but you'd need some really sophisticated audio equipment, and it'd take a hell of a long time to process the audio files generated. Still, with time... I can't wait 'til my computers at work have thumbprint scanners. ;)

I remember awhile back, seeing similar technology where they could recreate the contents of a monitor by the glow it makes on a wall, blinds, or window. It wasn't perfect, but they could piece together a word document at default font size and zoom; without ever seeing the monitor directly at all.

This is much cooler. (http://web.archive.org/web/20040209133232/tempest.nettwerked.net/research.html)

Back in the Army days part of our COMSEC was ensuring that typing couldn't be heard when inside vans and stuff. So not sure if this is just hitting mainstream or what but the military has known about it for a while and take measures to prevent it.

I worked with key encryption devices, that is to say if someone wanted to chat on our network they had to come see us for keys for their radios, and "other" devices. Was pretty neato stuff. But anyway I usually found myself in some soundproof van.

First read about somethinglike this in Neil Stephenson's excellent Cryptonomicon. Brings up some interesting legal questions. Say you're scanning a possible criminal in an adjacent hotel room, using a microphone or other receptor to pick up the sounds/signals of his typing on a laptop. You then convert the signal into text. While this is clearly eavesdropping, would/should it fall under wiretapping laws or is it just surveillance?

And how far will this kind of surveillance capability go? Soon they'll know what kind of porn you're looking at just from the fwap-fwap-fwap sound you're making.