Privacy-Preserving Gaze Data Streaming in Immersive Interactive Virtual Reality: Robustness and User Experience.

Area of Science:

• Computer Science
• Human-Computer Interaction
• Cybersecurity

Background:

• Eye tracking is increasingly integrated into virtual reality (VR) systems.
• Existing privacy research for eye tracking data is limited to data-centric utility metrics and black-box threat models.
• Re-identification attacks pose a significant privacy risk to users of VR systems.

Purpose of the Study:

• To propose and evaluate a methodology for assessing real-time privacy mechanisms in interactive VR applications.
• To incorporate user-centric utility metrics, including subjective user experience and task performance.
• To analyze the effectiveness of privacy mechanisms against various threat models.

Main Methods:

• Development of a novel methodology for evaluating real-time privacy mechanisms in VR.
• Incorporation of subjective user experience and task performance metrics into the evaluation.
• Assessment of privacy mechanisms against black-box, black-box with exemplars, and white-box threat scenarios.

Main Results:

• Re-identification accuracy was reduced to as low as 14% using evaluated privacy mechanisms.
• High usability scores and reasonable task performance were maintained.
• Differential effectiveness of privacy mechanisms against different threat models was identified.

Conclusions:

• The proposed methodology enables end-to-end assessment of re-identification risks in VR eye tracking.
• Effective privacy mechanisms can significantly mitigate re-identification risks while preserving user experience.
• This work advances the field of VR privacy by providing practical evaluation tools and insights.