The eye as a new side channel threat on smartphones

Abstract

Eye tracking is not a new idea in human-computer interaction research. Since at least as early as 1990s, researchers have tried to utilize eye movement to drive or monitor user interaction with computers. The new idea is using eye movement tracking to breach the privacy of mobile users. In this paper, we study the feasibility of exploiting consumer-grade cameras built onto current smartphones to log eye gazes, and then estimating the keypad numbers being tapped by the user. Assuming Trojan applications with camera use permissions, this process could be implemented without the user contest or knowledge, imposing a potential new threat to the security and privacy of mobile users. Our approach does not involve machine learning methods. In these first preliminary proof-of-concept experiments, we mainly rely on a human attacker to manually analyze the collected images from the smartphone. Utilizing basic dimensionality and motion flow calculations, our results show a promising attack vector with more than 60% of taps inference accuracy. © 2013 IEEE.