Behavio2Auth
Hebatollah Mostafa, Mohammad El‐Ramly, Abeer ElKorany, Hassan Shaban · 2019
Many mobile applications use mobile built-in sensors to provide users with a plethora of services to collect and store a mass of sensitive data. These mobile devices need to be protected from unauthorized access and allow access for legitimate users only. In this paper, the problem of unauthorized access is addressed by identifying the user during activities under two considered scenarios: walking and sitting.We authenticate users continuously and implicitly based on micro-movements by leveraging the typing activity information on the screen. These micro-movements come from the user's typing on the touchscreen while walking or sitting. Accelerometer data were analyzed to capture these micro-movements and build the proposed authentication model, named Behavio2Auth. Assuming that each individual has a distinct movement pattern, this hypothesis is used to differentiate between users. A set of experiments were conducted on 100 participants which show that Behavio2Auth is efficient for recognizing smartphone users during walking and sitting scenarios and achieves low Equal Error Rate of only (10%, 16%) with an Area Under the Curve of (95%, 91%).