Your Phone's Accelerometer Can Track Keystrokes On Your PC

Just when you thought you had all your security problems tied up, someone's gone and developed a way to interpret your keystrokes using nothing but a modern smartphone.

The US-based Georgia Institute of Technology is responsible for creating the technique, which started out based on microphones. The effectiveness of audio however pales in comparison to the extreme sensitivity (and convenience) of an accelerometer-equipped smartphone.

The iPhone 3GS's hardware wasn't good enough to execute the attack, but Patrick Traynor, an assistant professor at the university's School of Computer Science, says the added gyroscope in the iPhone 4 made a world of difference. The image above, provided by Georgia Tech, shows an Android device, so it's fair to say the technique would work on any phone with a sensitive enough gyroscope/accelerometer combination.

The attack attempts to predict keystrokes in pairs, using the distance between keys and their position on the keyboard (left versus right) as hints for a custom dictionary. As long as the word is longer than two letters, the system has a good chance of detecting what's been pressed. Here's a practical explanation from Georgia Tech's official release:

For example, take the word "canoe", which when typed breaks down into four keystroke pairs: "C-A, A-N, N-O and O-E." Those pairs then translate into the detection system’s code as follows: Left-Left-Near, Left-Right-Far, Right-Right-Far and Right-Left-Far, or LLN-LRF-RRF-RLF. This code is then compared to the preloaded dictionary and yields "canoe" as the statistically probable typed word.

The accuracy is at best 80 per cent, so your passwords aren't at risk. The effective range is also 7cm, which limits its usefulness. Still, many people leave their phones next to their notebooks, so all it could take is one wrong download to compromise your system.

But we're reaching here -- even Traynor believes its a bit of stretch, saying the odds of it being used in the wild to steal passwords are low.

Image: Georgia Tech

[Georgia Tech via Ars Technica]