This paper introduces a bi-modal mechanism that leverages the way a smartphone user signs on the touchscreen and taps/enters any ”text-independent” 8-digit numbers to authenticate their identity. Pre- cisely, by extracting the trajectory of touch-points and touch-timing features during the enrollment stage, our scheme creates a digital identity of a user based on these behaviors. In the verification stage, our scheme compares the captured touch-points and touch-timing signatures with the digital identity of the user created during the enrollment stage. If the captured signatures match the digital identity within a certain tolerance, the user is authenticated. The choice of low-level events, such as signing on the screen and touch-typing, as biometric modalities makes our scheme easier to implement and adapt. We evaluated our approach using multiple classifiers, i.e., K-Nearest Neighbor, Support Vector Machine, and Deep Neural Network, and achieved a high True Acceptance Rate of 97.1% with a low False Acceptance Rate of just 0.2%, and an accuracy of 98.45% on a dataset of 20 volunteers. These results prove our scheme accurate in verifying the identity of users while also maintaining a low rate of false acceptance of unauthorized users.
2IN1: A Bimodal Behavioral Biometric-based User Authentication Scheme for Smartphones
Attaullah Buriro;Flaminia Luccio
2024-01-01
Abstract
This paper introduces a bi-modal mechanism that leverages the way a smartphone user signs on the touchscreen and taps/enters any ”text-independent” 8-digit numbers to authenticate their identity. Pre- cisely, by extracting the trajectory of touch-points and touch-timing features during the enrollment stage, our scheme creates a digital identity of a user based on these behaviors. In the verification stage, our scheme compares the captured touch-points and touch-timing signatures with the digital identity of the user created during the enrollment stage. If the captured signatures match the digital identity within a certain tolerance, the user is authenticated. The choice of low-level events, such as signing on the screen and touch-typing, as biometric modalities makes our scheme easier to implement and adapt. We evaluated our approach using multiple classifiers, i.e., K-Nearest Neighbor, Support Vector Machine, and Deep Neural Network, and achieved a high True Acceptance Rate of 97.1% with a low False Acceptance Rate of just 0.2%, and an accuracy of 98.45% on a dataset of 20 volunteers. These results prove our scheme accurate in verifying the identity of users while also maintaining a low rate of false acceptance of unauthorized users.I documenti in ARCA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.