“Laser Bugging” the Passive Way: Recovering Audio from Video

Everyone knows the “laser bug” principle, right? This is idea that, by bouncing a laser beam off a window, you pick up the vibrations from inside the room — and you can now hear what people inside the room are saying.

Nowadays, of course, laser bugging doesn’t necessarily limit you to hearing speech. Any method that can get you room audio can (in theory) be used as a keylogger, since each key on your keyboard has a slightly different acoustic signature.

(In some cases, your computer will even transmit raw data acoustically, due to piezoelectricity and related phenomena in the components on your motherboard.)

The problem with all this is that lasers, while they can be made invisible to the naked eye, are still detectable. Someone bouncing a laser off your window risks pointing out to all and sundry, “HEY! EVERYONE WITH AN IR VIEWER! I’VE GOT A LISTENING POST OVER HERE WITH A BIG FRIGGIN’ LASER!”

The alternative to this, as recently pointed out by the NSA spy-gadgets catalog, is to bounce a microwave beam off some object inside the room. Traditionally this is a specially designed “retro-reflector bug” (such as the Great Seal Bug) but I’ve heard rumors the metal walls of e.g filing cabinets work too, given the right signal processing mojo.

Unfortunately, this is STILL DETECTABLE, given someone with a sensitive receiver and directional antenna. So what’s a spy to do?

Enter a brand new research trick from the folks at MIT.

By looking at the motion of objects captured on video, they’re able to recover the audio signal that caused those objects to move ever-so-slightly. This means that by pointing a high-speed camera at a bag of chips, a plant, or maybe even your curtains… they can recover (with poor fidelity) room audio.

In fact they can even recover audio (with even poorer fidelity) from regular video camera images, by taking advantage of the “rolling shutter” effect. (This effect means that horizontal movement at high speed results in a wavy vertical line — essentially a “record groove” in the picture.)

Now, unless you’re using a high speed video camera, their methods proved poor at recovering anything other than a very narrowband computer-audio track. But does that mean someone who’s going to take this idea and perfect it to a production-grade “field deployable” level will be unable to surmount the problem?

Probably best not to count on it… this is probably why the pros eschew windows.

%d bloggers like this: