[kefka]

Hmm. I never really thought about it until now. I used to be a hacker (not so nice one), but those days are long gone.

One idea with this is using a microphone array with HARK(http://www.hark.jp/). I would be able to listen to any arbitrary keyboard press, and map them to a 3d scene. Assuming a bit of jitter, I could probably reproduce what you typed on your keyboard.

A smartphone with touchscreen would be impractical with this setup. I'm unsure regarding buttoned smartphones (slide out keyboard).

I've been working with 3d scene generation and voice decoding. I'm making a board room auto-transcripter. It would map where people are, and attach who says what, when. It also has uses also in the courtroom where a mic array could also overhear whispers the jurors say in open court, to potentially catch issues that would cause a mistrial.

Of course, this could also be taken to the 'listen to everything in the area and decode semi-private actions'.

[3pt14159]

This is already an established attack vector. Furthermore, even if you have sound proofed the room high quality DSLRs can pick up enough vibration from things like plants to reenact the sound.

Edit: Also it goes beyond just password interception. By following patterns of key presses you can detect things like language and even application.

[EvilSockMonster]

There's already been a lot of research done on acoustic attacks. For example: http://www.inf.ed.ac.uk/publications/thesis/online/IM100855....

[skizm]

Would a mechanical keyboard be more vulnerable to that microphone type of attack since it is louder?

Not that many people are busting out a mechanical keyboard in a coffee shop. I'm just curious.