Continuous Smartphone User Authentication Based on Gesture-Sensor Fusion

With the widespread use of smartphones in daily life, ensuring the security of devices has become increasingly important. Traditional authentication methods, such as passwords and fingerprint recognition, typically only authenticate users once when unlocking the device, failing to provide continuous protection throughout a user's session. In contrast, continuous authentication utilizes behavioral biometric features to provide an imperceptible additional layer of protection for device security. However, current research often lacks universality, making it difficult to effectively address different scenarios, and the validation performance for different gestures is suboptimal. To overcome this limitation, in this paper, we present GSAuth, a continuous authentication scheme based on the fusion of sensors and gesture features. GSAuth combines measurements from motion sensors with tap and swipe gestures, creating two sets of feature sets, tap gesture feature set and swipe gesture feature set. A dynamic one-class classification model is designed to handle these diverse gesture feature vectors. Experimental evaluation is performed using data from 20 users in the open-source HMOG dataset. The results show that GSAuth performs well in processing datasets of tap gestures and swipe gestures, with Equal Error Rates (EER) of 0.074 and 0.108, respectively, which demonstrates the effectiveness and feasibility of GSAuth in practical applications.