Beyond recognition

Posted on 19 Jan 2007 at 16:31

 

Gallery

Technology now enables computers to identify you without you having to lift a finger, but will it protect your liberties - or undermine them? Jon Thompson considers how your movements are monitored with calm-computing analysis.

Your smartphone is peeking enticingly out of your pocket. An opportunistic thief picks your pocket, but before he's got 10 yards away, your phone is shouting "stop thief", dialling 999 and sending out a location detector signal. It then switches itself off, so the thief cannot access its contents.

Given the number of expensive gadgets we carry around with us, it would be comforting if we could rely on safety measures such as these should the worst happen. Several research groups say it's possible to identify a person from the way they type and surf the web, and even the way they walk - consolation for anyone who has been mugged for their mobile or laptop or fallen prey to an online identity thief. There is a darker side to this technology, though. It could also be used to monitor your movements, no matter how carefully you tried to hide your identity. In this feature, we take an in-depth look at this type of technology, how close it is to reality and just how accurate it might be.

Code comfort

Jani M?yj?i and colleagues at Finland's VTT Electronics in Oulu have been worried about the security of modern portable gadgets. "These devices contain increasing amounts of valuable personal information. For example, some smartphones contain wallet and e-commerce applications. Thus, the risk associated with them is increasing. Currently... sensitive data in mobile phones is protected with PIN codes. But in the 'on' state, not even the PIN code protects the information."

Leave your phone unlocked on a table in a caf?go to pay the bill and, by the time you return, someone else could have access to your life. However, M?yj?i believes he can make a phone identify who's carrying it. This idea uses some of the principles behind what's known as 'calm' computing. For an application to be considered calm, it mustn't demand action from the user. An application isn't considered calm when you need to take an action, such as entering a PIN or speaking a password for voice-recognition purposes.

GAIT expectations

A security system that recognises whether you're carrying it or not is an example of calm computing. This technology, called 'gait recognition', has long been used in computer vision systems. The VTT team wondered if they could directly capture and use the natural rhythmic motion a person generates when they walk. In theory, how someone walks relies on unique factors such as height, weight, bone and muscle structure, overall fitness and so on. As you walk, the position of your body changes rhythmically in 3D space.

Are these rhythms unique enough to identify someone? To find out, the team took an accelerometer and mounted it on a belt. A group of 19 men and 17 women then took turns to wear the belt and walk at slow, normal and brisk paces while it digitised and recorded the data generated by the accelerometer. This was the team's reference point. If the experiment was to work, software running on an associated laptop should be able to analyse future data and compare it against this reference, indicating who was subsequently wearing the device. A week later, the volunteers were asked to wear the belt again and walk while it captured more data. Results showed that the software could identify who was doing the walking.