PrivAR: Real-Time Privacy Protection for Location-Based Augmented Reality Applications

2025-08-04

AI Summary

To address location and trajectory privacy leakage caused by high-frequency GPS updates in location-based augmented reality (LB-AR) applications, this paper proposes the first client-side privacy-preserving framework tailored for real-time LB-AR. Our approach introduces two key innovations: (1) a Planar Staircase Mechanism (PSM), a differentially private location perturbation model achieving low positional error; and (2) a Thresholded Reporting PSM (TR-PSM), enabling selective location updates that jointly guarantee both point-wise and trajectory-level privacy. All noise injection and reporting decisions are performed on-device, ensuring strong privacy guarantees with minimal latency. Experimental results demonstrate that, compared to baseline methods, our framework improves game scores by 50%, increases adversary localization error by 1.8×, and incurs only a 0.06 ms increase in end-to-end latency—thereby simultaneously enhancing both QoS and privacy protection.

Technology Category

• Machine Learning: Privacy
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Intelligent Robots: Localization, Mapping, and Navigation

Application Category

• Security and Privacy: Privacy-enhancing technologies
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search

Abstract

Location-based augmented reality (LB-AR) applications, such as Pokémon Go, stream sub-second GPS updates to deliver responsive and immersive user experiences. However, this high-frequency location reporting introduces serious privacy risks. Protecting privacy in LB-AR is significantly more challenging than in traditional location-based services (LBS), as it demands real-time location protection with strong per-location and trajectory-level privacy guaranteed while maintaining low latency and high quality of service (QoS). Existing methods fail to meet these combined demands. To fill the gap, we present PrivAR, the first client-side privacy framework for real-time LB-AR. PrivAR introduces two lightweight mechanisms: (i) Planar Staircase Mechanism (PSM) which designs a staircase-shaped distribution to generate noisy location with strong per-location privacy and low expected error; and (ii) Thresholded Reporting with PSM (TR-PSM), a selective scheme that releases a noisy location update only when a displacement exceeds a private threshold, enabling many-to-one mappings for enhanced trace-level privacy while preserving high QoS. We present theoretical analysis, extensive experiments on two public datasets and our proprietary GeoTrace dataset, and validate PrivAR on a Pokémon-Go-style prototype. Results show PrivAR improves QoS (Gamescore) by up to 50%, while increasing attacker error by 1.8x over baseline with an additional 0.06 milliseconds runtime overhead.

Problem

• Real-time privacy protection for LB-AR applications
• Balancing low latency and high QoS with privacy
• Overcoming limitations of existing location privacy methods

Innovation

• Client-side privacy framework for real-time LB-AR
• Planar Staircase Mechanism for noisy location generation
• Thresholded Reporting with PSM for selective updates