Lightweight and Non-Invasive User Authentication on Earables.

The widespread adoption of wireless earbuds has advanced the developments in earable-based sensing in various domains like entertainment, human-computer interaction, and health monitoring. Recently, researchers have shown an increased interest in user authentication using earables. Despite the successes witnessed in acoustic probing and speech based authentication systems, this paper proposed a lightweight and non-invasive ambient sound based user authentication scheme. It employs the difference between the in-ear and out-ear sounds to estimate the individual-specific occluded ear canal transfer function (OECTF). Specifically, the {out-ear, in-ear} scaling factors at different frequency bands are captured via linear regression and treated as the OECTF for user authentication. The proposed system is validated using 12 subjects under six different noisy environments and achieves a Balanced Error Rate (BER) of 4.84%. The particularly lightweight system can be easily deployed in earbuds and paves the pathway for more personalized services.