New research from Lancaster University, Northwest University in China, and the University of Bath, which benefitted from funding from the Engineering and Physical Sciences Research Council (EPSRC), shows for the first time that attackers can crack Pattern Lock reliably within five attempts by using video and computer vision algorithm software.
By covertly videoing the owner drawing their Pattern Lock shape to unlock their device, while enjoying a coffee in a busy café for example, the attacker, who is pretending to play with their phone, can then use software to quickly track the owner’s fingertip movements relative to the position of the device. Within seconds the algorithm produces a small number of candidate patterns to access the Android phone or tablet.
The attack works even without the video footage being able to see any of the on-screen content, and regardless of the size of the screen. Results are accurate on video recorded on a mobile phone from up to two and a half metres away – and so attacks are more covert than shoulder-surfing. It also works reliably with footage recorded on a digital SLR camera at distances up to nine metres away.
Researchers evaluated the attack using 120 unique patterns collected from independent users. They were able to crack more than 95 per cent of patterns within five attempts.
Complex patterns, which use more lines between dots, are used by many to make it harder for observers to replicate. However, researchers found that these complex shapes were easier to crack because they help the fingertip algorithm to narrow down the possible options.
During tests, researchers were able to crack all but one of the patterns categorised as complex within the first attempt. They were able to successfully crack 87.5 per cent of median complex patterns and 60 per cent of simple patterns with the first attempt.