Boosting remote multi-user AR privacy through a magic rope

In remote Multi-user AR (MuAR), a user shares with remote users his/her physical environment, which is enhanced by computer-generated perceptual information. Despite its important role in Metaverse, a user may have serious privacy concerns for MuAR: the user may want to share only a portion, or may want to block certain areas in their environment. Today's object detection and AR tracking techniques fall short of reliably, efficiently, and accurately supporting this use case, in particular on an (already overloaded) headset without offloading to an edge/cloud server for privacy preservation purposes.

In this poster, we propose a novel primitive called Magic Rope to boost remote MuAR privacy. A user employs a flexible rope to circle enclosed areas as whitelisted (expose externally) or blacklisted (exclude/blur from sharing). The rope has specially designed markers allowing an AR headset to accurately and efficiently detect its enclosed area, as well as to drastically reduce the tracking overhead. We are working on addressing several research questions such as the marker design, efficient detection of chained markers on a rope, occlusion handling, multi-rope connection, and human-computer interaction design. We also plan to develop a full-fledged prototype of Magic Rope and integrate it with real remote MuAR applications. We will conduct extensive evaluations (including an IRB-approved user study) on real AR headsets.