Balancing the user privacy and expressiveness in VR mediated experiences

2021 – 2024


In the realm of Virtual Reality (VR), the advent of advanced sensors has facilitated the collection of user movement data with unprecedented accuracy. This development enables the creation of unique user profiles based on behavioural patterns, offering enhanced security and raising significant privacy concerns simultaneously. Our research investigates strategies to balance privacy and expressiveness in VR experiences.

First, we explored the challenges of hiding behavioural traits from machine learning algorithms designed for identity detection without using any utility technology. We investigated the impact of physical attributes, such as skin colour and facial hair, on the accuracy of these systems. We found them remarkably resilient, capable of identifying individuals accurately across multiple sessions and resistant to user attempts at deception.

Next, we investigated the effectiveness of computational behaviour filters to conceal users’ identities. Although data obfuscation methods effectively prevented algorithms trained on unfiltered data from identifying users, those trained on obfuscated data still managed to identify users with accuracy above random chance. As a solution, we developed three real-time behaviour filters that replace personal behaviours with generic expressions. These filters concentrate on facial expressions, eye movements, and hand gestures. Our findings show that these filters do not undermine the quality of interaction or the development of trust among collaborators, indicating significant progress in VR behavioural privacy.

To validate these behaviour filters, we devised a novel validation model pioneering in the VR privacy domain. Our research has progressed through the execution of four user studies involving adult participants, individually and in pairs, to assess the implications and effectiveness of these privacy issues and solutions in real-world scenarios.

As we continue our research, it becomes increasingly important to investigate behavioural privacy for VR users. Despite technological advancements, there is still a crucial need to educate users on the privacy risks associated with their behavioural data and the measures available to mitigate such risks. This research emphasizes the potential of behaviour filters in enhancing privacy and calls for increased awareness among users to protect their personal information in virtual environments.


    This project is funded by a grant from the Science for Technological Innovation National Science Challenge (SfTI).


    Researcher and Contact

    Dilshani Kumarapeli


    Supervisory Team

      • Rob Lindeman, HIT Lab NZ, University of Canterbury
      • Sungchul Jung, Software Development and Game Development, Kennesaw State University



      • Kumarapeli D, Jung S, and Lindeman RW. (2024). Privacy threats of behaviour identity detection in VR. Front. Virtual Real. 5:1197547. doi: 10.3389/frvir.2024.1197547
      • Kumarapeli D, Jung S, and Lindeman, RW. (2023). Privacy Threats of Behaviour Identity Detection in VR. 2023 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW), 861–862. IEEE. doi:10.1109/VRW58643.2023.00273
      • Kumarapeli D, Jung S, and Lindeman, RW. (2022). Emotional avatars: Effect of uncanniness in identifying emotions using avatar expressions. 2022 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW), 650–651. IEEE. doi:10.1109/VRW55335.2022.00176
      • Kumarapeli D. (2021). [DC] Privacy in VR: Empowering Users with Emotional Privacy from Verbal and Non-verbal Behavior of Their Avatars. 2021 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW), 715–716. IEEE. doi: 10.1109/VRW52623.2021.00240.


      Dilshani Kumarapeli, HIT Lab NZ

      Sungchul Jung, Kennesaw University