Behavioral biometrics in extended reality: implicit user authentication through artificial intelligence

With the growing proliferation of XR systems, and AR and VR applications, the demands on their security and data protection capabilities are also increasing. A central challenge lies in protecting XR devices from unauthorized access, particularly through reliable user identification. Conventional methods such as PINs or passwords have many limitations in XR contexts, as they are difficult to remember and cumbersome to enter in virtual environments. Moreover, they can be observed and compromised by outsiders who spy on an immersed virtual reality user. Therefore, there is a need for new, context-appropriate identification methods that can be seamlessly integrated into XR devices. One promising approach is the use of behavioral biometrics that are based on users’ body movements in XR. This method enables implicit and continuous user identification without requiring active input from the user, thereby offering a high degree of user-friendliness and security. This thesis investigates the suitability of behavioral biometrics through body movements for user identification in XR through six empirical studies. It analyzed movement components, physiological influencing factors, colocated interactions, and temporal changes in behavioral biometric patterns in XR. In addition, modalities such as gaze behavior, as well as head, hand, and finger movements associated with the interactions, were explored. The results show that behavioral biometric methods enable reliable identification in XR and that the recognition performance significantly depends on the analyzed factors and their specific characteristics.