eye movement biometric bias (undergrad thesis)

This work was completed as my undergraduate honors thesis with Dr. Eakta Jain at the University of Florida. Full text is coming soon pending publication in the UF Libraries collection!

Perceptual data inputs such as eye gaze enable rich interactions, especially in immersive virtual environments which increasingly integrate eye tracking. Despite the exciting new interaction modalities enabled by eye tracking, privacy concerns pervade and may be particularly relevant for groups whose identities are implicated by this perceptual data stream. This study explores the differences in identification rates and thus potential privacy risks for immersive eye-tracked virtual reality users with and without visual impairments, namely strabismus and amblyopia. Using the current state-of-the-art eye movement biometric algorithm on data collected during an eye-tracked VR game, we analyze the possibility of identifying users via gaze data. We find that users with visual impairments are identified at a higher rate (<73\%) than those without (<46\%) on the same duration of gaze data during the same task, and that this result pervades across data durations. These preliminary results elucidate inequities in privacy concerns with perceptual data collection and suggest that further exploration is needed to accurately represent the varied landscape of privacy concerns for diverse users.