Poster: Towards Privacy-Aware AR in a pervasive (unsafe) future

Abstract

Pervasive augmented reality earns its usefulness by continuously sensing the wearer's context, yet the founding pervasive-AR vision prices sensing only as benefit; every context source is also a leakage channel that can re-identify the wearer and captures bystanders who never consented. We recast privacy-aware context capture as a design space: a privacy-utility plane organized by three strategies—avoidance, mitigation, protection—evidenced by three systems. The representation of an already-permitted signal is itself a privacy control, reshaping the achievable frontier rather than sliding along it; and because privacy norms are contextual, context-aware AR should navigate this plane at runtime—on-device, by default, for everyone.