PrivateXR: Defending Privacy Attacks in Extended Reality Through Explainable AI-Guided Differential Privacy
🤖 AI Summary
In AI–XR integration, sensitive biometric data—such as eye-tracking signals—are vulnerable to membership inference attacks (MIA) and re-identification attacks (RDA), yet conventional differential privacy (DP) methods degrade model accuracy and increase latency due to global noise injection, failing to meet XR’s stringent real-time requirements.
Method: We propose an XAI-guided adaptive DP framework that dynamically couples explainable AI (XAI) techniques—specifically SHAP and LIME—with DP to enable feature-level, sensitivity-aware privacy budget allocation; only the most privacy-sensitive features are perturbed during inference, balancing privacy, accuracy, and low latency.
Contribution/Results: To our knowledge, this is the first DP system deployed in real time on edge XR hardware (HTC VIVE Pro) with user-controllable privacy levels. Evaluated on motion sickness prediction, it reduces MIA and RDA success rates by 43% and 39%, respectively, retains 97% model accuracy, achieves ~2× inference speedup, and demonstrates practical usability and high user satisfaction via empirical user studies.
Application Category
📝 Abstract
The convergence of artificial intelligence (AI) and extended reality (XR) technologies (AI XR) promises innovative applications across many domains. However, the sensitive nature of data (e.g., eyetracking) used in these systems raises significant privacy concerns, as adversaries can exploit these data and models to infer and leak personal information through membership inference attacks (MIA) and re-identification (RDA) with a high success rate. Researchers have proposed various techniques to mitigate such privacy attacks, including differential privacy (DP). However, AI XR datasets often contain numerous features, and applying DP uniformly can introduce unnecessary noise to less relevant features, degrade model accuracy, and increase inference time, limiting real-time XR deployment. Motivated by this, we propose a novel framework combining explainable AI (XAI) and DP-enabled privacy-preserving mechanisms to defend against privacy attacks. Specifically, we leverage post-hoc explanations to identify the most influential features in AI XR models and selectively apply DP to those features during inference. We evaluate our XAI-guided DP approach on three state-of-the-art AI XR models and three datasets: cybersickness, emotion, and activity classification. Our results show that the proposed method reduces MIA and RDA success rates by up to 43 % and 39 %, respectively, for cybersickness tasks while preserving model utility with up to 97 % accuracy using Transformer models. Furthermore, it improves inference time by up to $approx 2 imes$ compared to traditional DP approaches. To demonstrate practicality, we deploy the XAI-guided DP AI XR models on an HTC VIVE Pro headset and develop a user interface (UI), namely PrivateXR, allowing users to adjust privacy levels (e.g., low, medium, high) while receiving real-time task predictions, protecting user privacy during XR gameplay. Finally, we validate our approach through a user study, which confirms that participants found the PrivateXR UI effective, with satisfactory utility and user experience.
Problem
Research questions and friction points this paper is trying to address.
Defends privacy attacks in XR using explainable AI and differential privacy.
Selectively applies DP to key features to maintain model accuracy and speed.
Reduces membership inference and re-identification attacks while preserving real-time utility.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Explainable AI guides differential privacy feature selection
Selective DP application reduces noise and preserves model accuracy
Real-time privacy adjustment in XR via user interface PrivateXR
🔎 Similar Papers
No similar papers found.
