Smart glasses pose significant privacy challenges by capturing bystanders without notice or consent. Existing solutions often rely on permanent obfuscation, shift responsibility to bystanders, or offload sensitive data to the cloud, risking unauthorized access and denying meaningful control. We present a novel three-tier architecture for privacy-preserving smart glasses that enforces blurring at the point of capture, supports synthetic face replacement, and enables consent-based decryption of visual data. We implement SITARA, a working prototype on Raspberry Pi–based hardware, and demonstrate on-device blurring and secure consent mediation. Our evaluation shows that SITARA operates efficiently while achieving reliable bystander anonymization; furthermore, its synthetic replacement delivers perceptual quality competitive with state-of-the-art baselines, all without exposing raw video or undermining usability.
IEEE International Conference on Pervasive Computing and Communications (PerCom)
2026-03-20
2026-06-10