BioMove: Biometric User Identification from Human Kinesiological Movements for Virtual Reality Systems

Ilesanmi Olade, Charles Fleming, Hai-Ning LiangSensors 2020 · OpenReview

TL;DR

This paper introduces BioMove, a novel system for continuous, non-intrusive user identification in Virtual Reality (VR) by analyzing kinesiological movements (head, eye, and hand). BioMove achieves high accuracy (98.6%) and demonstrates strong resistance to simulated attacks, making VR environments more secure and personalized.

What Problem Does This Paper Address?

As Virtual Reality (VR) becomes more prevalent in entertainment, education, and business, the need for secure and seamless user identification grows. Traditional authentication methods like PINs or passwords are cumbersome and vulnerable to observation in VR. Furthermore, the unique, immersive nature of VR presents new security challenges and opportunities. This paper addresses the gap in robust, transparent, and non-intrusive identification mechanisms specifically tailored for VR environments, leveraging the inherent uniqueness of individual human movement patterns.

Methods

The core of BioMove lies in capturing and analyzing a user's kinesiological data within a VR environment. The researchers designed a VR setup where participants performed specific tasks involving grabbing, rotating, and dropping objects. This allowed them to collect synchronized data from:

  • Head-Mounted Display (HMD): Capturing 6-DOF (six degrees of freedom) head movements (position and rotation).
  • Eye Tracker: Recording gaze positional data to understand where the user is looking.
  • Hand-Held Controller (HHC): Tracking 6-DOF hand movements (position and rotation).

This multi-sensor data was then processed through a machine learning pipeline. Specifically, they employed a k-Nearest Neighbors (kNN) classifier (with k=10) for both task identification and participant identification. To ensure consistency across participants and varying task completion times, a motion data resampling technique was used to standardize the number of movement vectors per task. The system was trained and tested using an 80/20 split of the collected data.

Datasets

The study utilized a custom dataset collected from 15 participants performing a series of controlled tasks within a specifically designed VR environment. The dataset comprises synchronized head, eye, and hand movement data captured at 25 Hz. No pre-existing public datasets were used for training or evaluation.

Key Insights and Ideas

  • Kinesiology as a Biometric: The fundamental insight is that individual kinesiological patterns are unique and can serve as a reliable biometric for identification, especially in the context of VR where movement is central to interaction.
  • Multi-Sensor Fusion: Combining data from head, eye, and hand movements provides a richer and more discriminative feature set than relying on a single source, enhancing accuracy and robustness.
  • Continuous and Transparent Identification: BioMove enables identification without explicit user action, making it seamless and non-intrusive, which is crucial for user experience in VR.
  • VR-Specific Threat Model: The paper considers unique VR threats, such as attackers with similar physical features, and demonstrates how BioMove can counter them.
  • Task-Specific Identification Models: Training separate identification models for different tasks can improve accuracy and response time by narrowing down the search space.

Key Results

  • High Classification Accuracy: The BioMove system achieved a cross-validated classification accuracy of 98.6% for user identification.
  • Fast Processing Speed: The identification processing speed was remarkably fast, averaging 0.035 ms per classification, making it suitable for real-time applications.
  • Robustness Against Attacks: In a whitebox penetration test with 12 attackers, none achieved a confidence value above 50%. Attackers with physically similar features to the valid users showed higher confidence values but still remained below the security threshold.
  • Low Error Rate: The system demonstrated a low error rate of 1.4%.

Limitations and Future Work

  • Demographic Limitations: The study primarily recruited participants between 18 and 30 years old. The authors acknowledge that the effect of aging on kinesiological movements and its impact on the system's accuracy needs further investigation.
  • Integration with VR Frameworks: Future work aims to integrate BioMove into existing VR environment frameworks and firmware, enabling it to function as a universal identification module across various VR applications.
  • Authentication Criteria: While the paper focuses on identification, extending the system to provide robust authentication (i.e., rejecting invalid users with high confidence) requires further development of rejection criteria.
  • Broader Task Variety: The current study used a limited set of tasks. Exploring a wider range of VR activities and their associated kinesiological patterns would further validate the system's generalizability.

Summary updated

Abstract

Virtual reality (VR) has advanced rapidly and is used for many entertainment and business purposes. The need for secure, transparent and non-intrusive identification mechanisms is important to facilitate users’ safe participation and secure experience. People are kinesiologically unique, having individual behavioral and movement characteristics, which can be leveraged and used in security sensitive VR applications to compensate for users’ inability to detect potential observational attackers in the physical world. Additionally, such method of identification using a user’s kinesiological data is valuable in common scenarios where multiple users simultaneously participate in a VR environment. In this paper, we present a user study (n = 15) where our participants performed a series of controlled tasks that require physical movements (such as grabbing, rotating and dropping) that could be decomposed into unique kinesiological patterns while we monitored and captured their hand, head and eye gaze data within the VR environment. We present an analysis of the data and show that these data can be used as a biometric discriminant of high confidence using machine learning classification methods such as kNN or SVM, thereby adding a layer of security in terms of identification or dynamically adapting the VR environment to the users’ preferences. We also performed a whitebox penetration testing with 12 attackers, some of whom were physically similar to the participants. We could obtain an average identification confidence value of 0.98 from the actual participants’ test data after the initial study and also a trained model classification accuracy of 98.6%. Penetration testing indicated all attackers resulted in confidence values of less than 50% (<50%), although physically similar attackers had higher confidence values. These findings can help the design and development of secure VR systems.
BibTeX
@inproceedings{olade2020biomove,
  author = {Ilesanmi Olade and Charles Fleming and Hai-Ning Liang},
  title = {{BioMove: Biometric User Identification from Human Kinesiological Movements for Virtual Reality Systems}},
  booktitle = {Sensors},
  year = {2020},
  url = {https://openreview.net/forum?id=qEewBdGWWn},
}
Figure