Be careful *how* you type ...

[Rakido]

http://www.computerworld.com.au/index.php/id;1176568408;fp;2;fpid;1

Researchers at the University of California, Berkeley, have found a way to turn the clicks and clacks of typing on a computer keyboard into a startlingly accurate transcript of what exactly is being typed.

In a paper released last week, the researchers explained how they developed software that could analyze the sound of someone typing on a keyboard for just 10 minutes and then piece together as much as 96 percent of what had been typed.

[Jammrock]

some genious figured out a way to translate keystrokes by hooking a torroid and a transmitter to the keyboard cable. Sends every keyboard stroke to a remote location

SPying is fun ... despite being highly illegal. Maybe that's what makes it fun?

[cjolley]

Or, you could produce the world's lowest cost wireless keyboard.
Just set your computer's mic next to a keypad with no electronics in it.

[Gurm]

I challenge them to actually make this work for me without being able to see what I've typed.

I could see it working for "most" typists, IF they can calibrate it first to the material that you're typing. But working double-blind? Not gonna happen.

And even if you know what I've typed in the past, the odds of figuring it out from me? Practically nil. Ask Damian - he's seen me type.

[Wombat]

I challenge them to actually make this work for me without being able to see what I've typed.

I could see it working for "most" typists, IF they can calibrate it first to the material that you're typing. But working double-blind? Not gonna happen.

And even if you know what I've typed in the past, the odds of figuring it out from me? Practically nil. Ask Damian - he's seen me type.

There's a decent amount that can be done. 10 minutes of coherent typing should be more than enough to do some statistical analysis. You can make assumptions about letter frequency with that amount of time, and I'd be surprised if the space bar weren't a distinct sound.

[Gurm]

There's a decent amount that can be done. 10 minutes of coherent typing should be more than enough to do some statistical analysis. You can make assumptions about letter frequency with that amount of time, and I'd be surprised if the space bar weren't a distinct sound.

Sure sure. R S T L N E. Beyond that? Not so much.

And yeah, the space bar makes a distinctive sound.

But the rest of the keys, especially on one of these UltraX keyboards, are pretty much just consistent little chickle sounds.

[Jammrock]

The software probably measuers things like the time between keystrokes, the position of the sound in a three-dimensional space, the number of clicks between a spacebar stroke (a very distinct sound) and things like that. Of course, if you're IM'ing and do it like Gurm and I, you hit the enter key every 6-7 words...

SO your sentences
start to look like this
because in IM'ing you have to
things differently that normal
just to drive people absolutely crazy
that you are typing incoherently.

Jammrock

[DGhost]

assuming that the space bar makes the most distinct sound and you are capable of seperating words out, it wouldn't be too hard to run the statistically probably cases across a word list and come up with another list of what is likely and what is not.

this goes in the basket with being able to read data from the lights on routers.

[Claymonkey]

I HAVE seen Gurm type but I'd be willing to bet that if you can get this to work on more than one person in the first place you can get it to work on most anyone... including gurm. Quite laptop style keys or not it's going to make SOME sound.

Still 96% accuracy is still terrible. This is why speech recognition and/or OCR software doesn't do well even if you have 98-99% accuracy. 1% of inaccuracy can send a document into tailspin of illegibility. You take out the human element to clean it up and it's unprofessional and that terrible looking word document sent to the wrong person can get you fired.

[Wombat]

But the rest of the keys, especially on one of these UltraX keyboards, are pretty much just consistent little chickle sounds.

And their timing matters a good bit. You can't type two keys with the same finger very quickly. If their mics and algorithms are very good, they may even be able to sense the amount of occlusion from the hands. Timing, plus common letter combinations would yield a lot of data.

[Gurm]

I still doubt that it's real. This ranks right up there, for me, with the "we can see through walls using multiple tomography" crap.