Can mobile device users be identified by how they play a game?

A consumer reports survey says that 34% of smart phone users take no security precautions with their devices [1], in spite of the various security features available now-a-days on smartphones. This may be due to usability problems posed by the current authentication systems on a mobile device. The aim of our project is to develop more usable authentication for mobile devices by studying and modeling the manner in which users interact with games. To assess the viability of authenticating users based on gameplay we performed an IRB approved experiment with 12 human participants. Our subjects played 3 popular commercially available games on an Android device while we recorded their touchscreen usage. A Support Vector Machine was then used to classify users by modeling 19 touch screen behavioral features. Our modeling resulted in at least a 90% true positive rate for each of the 3 games while keeping the false reject rate below 1%. These results provide initial evidence that game play elements lead the users to behave in unique manner thus providing us with distinguishable touch screen features.