EarCapAuth: Biometric Method for Earables Using Capacitive Sensing Eartips

TL;DR

EarCapAuth, an authentication mechanism using 48 capacitive electrodes embedded into the soft silicone eartips of two earables, outperforms some earable biometric principles from related work.

Abstract

Earphones can give access to sensitive information via voice assistants which demands security methods that prevent unauthorized use. Therefore, we developed EarCapAuth, an authentication mechanism using 48 capacitive electrodes embedded into the soft silicone eartips of two earables. For evaluation, we gathered capactive ear canal measurements from 20 participants in 20 wearing sessions (12 at rest, 8 while walking). A per user classifier trained for authentication achieves an EER of 7.62% and can be tuned to a FAR (False Acceptance Rate) of 1% at FRR (False Rejection Rate) of 16.14%. For identification, EarCapAuth achieves 89.95%. This outperforms some earable biometric principles from related work. Performance under motion slightly decreased to 9.76% EER for authentication and 86.40% accuracy for identification. Enrollment can be performed rapidly with multiple short earpiece insertions and a biometric decision is made every 0.33s. In the future, EarCapAuth could be integrated into high-resolution brain sensing electrode tips.

Figures (4)

• Figure 1: The custom-built device used to realise the EarCapAuth authentication method. Each earbud (left or right) contains the two MPR121 capacitive sensor connected to 8 electrodes each. The flex printed-circuit board (flex PCB) is folded and glued onto a soft silicone eartip. A 3D-printed enclosure houses the components. The earpiece is connected via four cables to the JST connector of OpenEarable oe which records capacitive sensing data to an on-board microSD card.
• Figure 2: The left ear (L1-L24) and right ear (R1-R24), capacitive signals are used to generate a raw data frames as 1D vectors at 15 Hz. For authentication a linear support vector machine (SVM) makes an accept or reject decision for user authentication based on the arithmetic mean of five samples frames. The identification model also uses a linear SVM as classifier. It predicts the respective user from the dataset (User 1, User 2, User 3, etc.).
• Figure 3: The heatmaps each show the mean of 48 capacitive electrode readings of one chunk (arithmetic mean of 5 samples per channel). They were selected randomly from three different wearing sessions for every participant. The heatmaps show the consistency within participants and the variation between different participants. The color scale represents the electrode activity, with values normalized across all participants and sessions shown in the figure.
• Figure 4: (A) Changing the acceptance threshold value allows fine tuning the model to allow higher false acceptance or false rejection rates; (B) Confusion matrix of one fold of the identification task for 20 participants shows high accuracy for most participants with consistent confusions between participants and wearing sessions; (C) Identification accuracy improves with more training.