Continuous Authentication in VR Environments Using LSTM: An Experimental Study

Conventional PIN (Personal Identification Number) and password-based authentication methods face significant limitations in virtual reality (VR) environments, including challenges with memorization and susceptibility to eavesdropping. Simple PINs and passwords compromise authentication security, while biometric authentication, although promising, lacks the ability to provide continuous verification without active user involvement. The proposed work introduces a continuous authentication system for VR environments that leverages head motion as a biometric characteristic. To address privacy concerns associated with outsourcing motion sensor data to servers for training purposes, the proposed system encrypts sensor data before transmission, mitigating the risk of sensitive information, such as user PINs or passwords, being inferred. The system employs long-short-term memory (LSTM) networks to classify head motion patterns, achieving high average accuracy of 99% in distinguishing between authorized and unauthorized users in the two raw datasets without encryption and around 97% using encrypted motion sensor data for user authentication Furthermore, the solution seamlessly integrates with existing VR hardware, requiring no OS-level modifications or additional components, offering a secure, practical, and efficient authentication mechanism for immersive VR technologies.