MIRAGE: Full-Body Bystander Privacy for Smart Glasses with Consent-Based Restoration
MIRAGE masks full-body biometrics on smart glasses and restores footage only with consent.
MIRAGE is a three-tier privacy system for smart glasses that conceals full-body biometric cues, including gait, posture, and silhouette, while allowing synthetic body replacement and encrypted consent-based restoration. It was implemented on a Raspberry Pi 5, companion phones, and a cloud generative backend. Visible-body detection reaches 0.948 average precision and 0.976 average recall. Bounding-box masking reduces silhouette re-identification to 10.86% Rank-1 versus an 11.12% chance level, and adaptive gait identification falls from 90.25% to 26.20% Rank-1.
- Hides gait, posture, and silhouette, not only faces.
- Body detection reaches 0.948 AP and 0.976 AR.
- Silhouette re-identification falls to about chance level.
- Adaptive gait Rank-1 drops from 90.25% to 26.20%.
Full article143 words · extracted from arxiv.org · click to collapse
Video recording on smart glasses exposes more than faces. Continuous capture reveals full-body biometric signatures, including gait, posture, and silhouette, that enable person re-identification (ReID) even after conventional face sanitization. We present MIRAGE, a three-tier architecture for privacy-preserving smart glasses that enforces full-body privacy, supports synthetic full-body replacement, and retains encrypted recovery material for consent-based restoration. We implement MIRAGE on a Raspberry Pi~5 (a CPU-only proxy for smart-glasses compute), companion phones, and a cloud generative backend. Compared to prior systems, MIRAGE achieves 0.948 AP and 0.976 AR while accurately detecting the complete visible body. Its bounding box masking reduces learned silhouette-based ReID to essentially random guessing, with 10.86% Rank-1 accuracy compared with an 11.12% measured chance level. Even against an adaptive adversary retrained on MIRAGE's sanitized pose signals, Rank-1 gait identification drops from 90.25% to 26.20%, removing 72.5% of the adversary's identification advantage.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.24537