Adaptive Threshold Scheme for Touchscreen Gesture Continuous Authentication Using Sensor Trust

In this study we produce a continuous authentication scheme for mobile devices that adjusts an adaptive threshold for touchscreen interactions based on trust in passively collected sensor data. Our framework unobtrusively compares real-time sensor data of a user to historic data and adjusts a trust parameter based on the similarity. We show that the trust parameter can be used to adjust an adaptive threshold in continuous authentication schemes. The framework passively models temporal, spatial and activity scenarios using sensor data such as location, surrounding devices, wi-fi networks, ambient noise, movements, user activity, ambient light, proximity to objects and atmospheric pressure from study participants. Deviations from the models increases the level of threat the device perceives from the scenario. We also model the user touchscreen interactions. The touchscreen interactions are authenticated against a threshold that is continually adjusted based on the perceived trust. This scheme provides greater nuance between security and usability, enabling more refined decisions. We present our novel framework and threshold adjustment criteria and validate our framework on two state-of-the-art sensor datasets. Our framework more than halves the false acceptance and false rejection rates of a static threshold system.