ParaSight: Enabling Privacy-preserving Sensing Data Sharing via Device-to-device Utterance-based Communication

Youngwook Do, Seong-Eun Moon, Minsuk Chang

Abstract: Privacy-invasive sensors such as cameras and microphones in smart devices (e.g., Facebook Portal, Google Nest Hub Max, Amazon Echo Show) are now ubiquitous in a user’s private setting, which raises users’ privacy concerns of always listening and monitoring. To address such privacy concerns, we developed ParaSight, a smart speaker add-on device that communicates to users sensing data transfer. ParaSight transmits the information yet speaks out loud what is being transmitted to a smart speaker connected to a server via utterances by leveraging the smart speaker’s ability to understand human languages. Communicating sensing data as utterances brings two benefits: (i) a user can hear and understand what data, and when, are being transferred as utterances are human-understandable; (ii) raw data are processed locally and only filtered signals can be uploaded to a server. ParaSight, for example, can receive and locally process raw data of a user’s audio and video data (e.g., snoring sound, workout video), and only transmit the filtered data to a server. We demonstrate two work-in-progress applications—snoring detection and home training applications—to show ParaSight’s use cases.