Dissertation Defense: Shwetha Rajaram

The School of Information is pleased to announce the oral defense of Shwetha Rajaram.

Title: Privacy-Driven Adaptation of Augmented Reality Interfaces

Date: Thursday, June 4, 2026

Time: 1:00-3:00pm

Location: LCSIB 2290 and Online via Zoom

Zoom Meeting ID: 983 9029 5518; Passcode: 347892

Michael Nebeling, serving as committee chair, will preside over the oral defense. 

All are welcome to attend!

Abstract:

As augmented reality (AR) glasses become increasingly lightweight and intelligent, they are emerging as powerful tools to enhance everyday workflows. Equipped with environmental and biometric sensing capabilities, AR glasses analyze users' activities and superimpose digital content on top of their physical surroundings, providing in-situ assistance for learning, collaboration, accessibility, and beyond. However, these capabilities raise significant privacy concerns: AR interaction data can reveal people's identities, activity patterns, and health conditions; spatial maps can capture sensitive environmental details and bystanders, without their awareness or consent. These risks are compounded by a lack of effective privacy controls for AR, as existing solutions often degrade usability or core functionality, and by users' limited capacity to actively manage privacy in dynamic settings.

To mitigate privacy risks across everyday contexts, this dissertation proposes and investigates privacy-driven AR adaptation: a process that balances user experience (UX) and privacy across AR users and bystanders by adjusting interactions in a fine-grained manner.

This thesis explores privacy-driven AR adaptation in three parts, contributing a conceptual foundation alongside design and technical interventions across the AR development and usage lifecycle. Part 1 systematically establishes a design space of AR interaction techniques that enable core functionality while reducing data exposure, derived from scenario-based elicitation studies with 10 AR researchers. To make this design space actionable for developers, it embeds the techniques into a visualization tool, validated through workshops with 6 AR developers. Then, Part 2 guides the creation of such privacy-friendly AR interactions for designers and developers without formal privacy training. It first scaffolds the elicitation method with a threat modeling task; a comparative study demonstrates how this helps developers generate creative and privacy-friendly interactions. It then develops an AR storyboarding tool that raises designers' awareness of potential privacy risks by visualizing threat representations within their prototypes. Finally, Part 3 automates AR interface adaptations to safeguard privacy with minimal user effort. It formulates an adaptation model that optimizes AR sensing permissions to balance UX and privacy for multiple users and bystanders. A simulation and analysis toolkit is developed to evaluate this model and serve as a platform for future research; studies with 8 security & privacy researchers demonstrate its utility.