Mizzou Engineering Research Highlights Security and Privacy Risks of AI in Extended Reality (XR)

At the University of Missouri (Mizzou) College of Engineering, researchers are tackling the growing cybersecurity and privacy challenges arising from the fusion of artificial intelligence (AI) and extended reality (XR) — immersive technologies combining virtual and augmented reality.

Led by Dr. Khaza Anuarul Hoque, assistant professor in the Department of Electrical Engineering and Computer Science, the Dependable Cyber-Physical Systems Lab is investigating how AI-enhanced XR applications in defense, healthcare, education, and industry can both transform and endanger user experiences.

Key Findings and Innovations:

User Data Vulnerability: XR systems collect sensitive data (e.g., heart rate, gaze tracking), which can be exploited through network vulnerabilities for identity theft, surveillance, or manipulation.

AI-Powered Threat Prediction: Hoque’s team used explainable AI and physiological datasets to predict cybersickness and cognitive overload in VR users with over 90% accuracy.

Few-Shot Learning Advances: A second study introduced a lightweight AI training approach to predict cybersickness, offering rapid, user-specific tuning ideal for large-scale XR deployments.

Defense Against Membership Inference Attacks: Using differential privacy, the team protected users’ data without degrading AI performance, reducing attack success rates by over 50%.

Supported by the Department of Defense (DoD) and National Science Foundation (NSF), the lab’s work has been featured at major conferences such as IEEE VR and ISMAR 2024.

Dr. Hoque emphasizes that with the rapid evolution of large language and vision models, new threats continue to emerge. “It’s a never-ending race,” he said, “but at Mizzou Engineering, we’re staying one step ahead.”

🔗 Source: Mizzou Engineering