Who’s Listening? Analyzing Privacy Preferences in Multi-User Smart Personal Assistants Settings

Authors: 

Carolina Carreira, Carnegie Mellon University, IST University of Lisbon, INESC-ID; Cody Berger, Khushi Shah, Samridhi Agarwal, Yashasvi Thakur, McKenna McCall, Nicolas Christin, and Lorrie Faith Cranor, Carnegie Mellon University

Abstract: 

Smart personal assistants (SPAs) are voice-activated devices that help users with daily tasks such as setting alarms and controlling other smart devices. SPAs' voice recognition capabilities allow them to respond differently to different users. Despite this, privacy controls for these devices are typically coarse-grained, offering little flexibility for individualized preferences. This means devices shared by several users may not be able to meet all of their privacy needs simultaneously. To understand whether privacy settings available for today's SPAs meet different user groups' privacy preferences, we conducted a 90-participant survey and an expert evaluation of the privacy settings for two popular SPAs. Primary and secondary users of SPAs seem to have higher privacy preference acceptability for closer relationships like partners and children accessing data like voice recordings and activity history, compared to more distant actors like neighbors or advertising agencies. However, even within closer circles, users prefer retaining control over changing certain privacy settings rather than fully delegating that control. Our qualitative results reinforce these findings, with common concerns around unauthorized information sharing, audio monitoring, and data breaches from outside agents. Our results highlight the need for flexible, granular privacy controls to adapt to users' diverse preferences across different relationships and contexts.

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.