Sorry, Apple. The fingerprint recognition feature on the
upcoming iPhone 5s, Touch ID, might be eye-catching, but you still have to log
into your device. Identifying someone by the way they tap and swipe on a
touchscreen might be the more natural, unobtrusive future of smartphone
biometricsMovie Camera.
Developed by Cheng Bo at the Illinois Institute of
Technology and his colleagues, SilentSense does just that. Using the phone's
built-in sensors, it records the unique patterns of pressure, duration and
fingertip size and position each user exhibits when interacting with their
phone or tablet.
Machine learning algorithms then turn this into a signature
that identifies the user – and will lock out anyone whose usage patterns do not
match.
To increase the system's accuracy, the smartphone's
accelerometer and gyroscope measure how much the screen moves when you are
jabbing at it. They can also pick up on your unique gait as you walk while
using the screen.
"Different users, dependent on sex and age among other
things, will have different habits in interacting," says Bo.
In tests, 100 users were told to use the smartphone's
touchscreen as they would normally. SilentSense was able to identify the
phone's owner with 99 per cent accuracy after no more than 10 taps. Even with
an average of 2.3 touches the system was able to verify the user 98 per cent of
the time.
To save on power, the software stops checking the user's
identity when apps like games are being used. To maintain security, it
automatically switches on when more sensitive applications, such as email or
SMS, are accessed.
"This is interesting, creative research," says
Kevin Bowyer, a biometrics researcher at the University of Notre Dame, Indiana.
"It could help preserve privacy if the phone could recognise if the owner
or a guest was using it and lock guests out of some applications."
Journal reference: arXiv:1309.0073v1