Motion ID: Gesture-Based Biometrics for User Identification and Authentication in Virtual Reality

This study explores a novel approach to gesturebased biometrics for identifying and authenticating virtual reality (VR) users, focusing on methods beyond traditional task-based systems to improve real-world applicability. By employing a Random Forest model, the research assesses various gesturessuch as hand waves, finger wiggles, and YMCA dance movements-analyzing accuracy across different VR headsets. These models are both able to identify which user out of a user group is using the headset from the motion data as well as authenticating them based on confidence levels. This cross-headset evaluation, using a sliding window technique with Euclidean distance for joint-based tracking, establishes a more reliable means of user authentication that surpasses prior studies in accuracy. By using shorter data segments ($\mathbf{1 - 2. 5}$ seconds) for identification, this approach enables faster and more practical authentication along with smaller training data segments for model creation. Results suggest that gesture biometrics could enhance fields such as multi-factor authentication and continuous biomarker tracking. This work contributes to existing VR biometrics literature by demonstrating that gesture-based biometrics can effectively authenticate users across diverse VR devices, potentially leading to more secure and user-friendly applications.