AccLock: Unlocking Identity with Heartbeat Using In-Ear Accelerometers

TL;DR

AccLock is a passive, unobtrusive earphone-based authentication system that uses in-ear heartbeat signals and advanced deep learning techniques to verify users securely and reliably without active user involvement.

Contribution

The paper introduces a novel passive authentication system leveraging in-ear heartbeat signals, with a new denoising scheme and a disentanglement-based deep learning model for scalable user verification.

Findings

Achieved an average FAR of 3.13% and FRR of 2.99% in experiments with 33 participants.

Developed a two-stage denoising scheme to improve signal quality in in-ear BCG signals.

Implemented a Siamese network-based framework that scales without per-user classifier training.

Abstract

The widespread use of earphones has enabled various sensing applications, including activity recognition, health monitoring, and context-aware computing. Among these, earphone-based user authentication has become a key technique by leveraging unique biometric features. However, existing earphone-based authentication systems face key limitations: they either require explicit user interaction or active speaker output, or suffer from poor accessibility and vulnerability to environmental noise, which hinders large-scale deployment. In this paper, we propose a passive authentication system, called AccLock, which leverages distinctive features extracted from in-ear BCG signals to enable secure and unobtrusive user verification. Our system offers several advantages over previous systems, including zero-involvement for both the device and the user, ubiquitous, and resilient to environmental noise. To realize this, we first design a two-stage denoising scheme to suppress both inherent and sporadic interference. To extract user-specific features, we then propose a disentanglement-based deep learning model, HIDNet, which explicitly separates user-specific features from shared nuisance components. Lastly, we develop a scalable authentication framework based on a Siamese network that eliminates the need for per-user classifier training. We conduct extensive experiments with 33 participants, achieving an average FAR of 3.13% and FRR of 2.99%, which demonstrates the practical feasibility of AccLock.