Exploring Reliable PPG Authentication on Smartwatches in Daily Scenarios

TL;DR

The paper tackles the challenge of reliable, low-power PPG-based authentication on smartwatches in daily life, where motion and physiological variability degrade performance. It introduces MTL-RAPID, a compact multi-task model that jointly evaluates waveform quality and performs identity verification through the RAPID architecture, achieving state-of-the-art AUCs up to 99.2% and EER as low as 3.5% under challenging conditions. The authors validate the approach across daily, cross-day, and usability studies, demonstrate efficiency with 80k parameters and sub-2 ms inference, and open-source a comprehensive wrist-PGG dataset (ANT) to foster reproducible research. The work highlights the potential for unobtrusive, secure smartwatch authentication while outlining future directions like adaptive learning and multi-modal fusion to further address long-term variability.

Abstract

Photoplethysmography (PPG) Sensors, widely deployed in smartwatches, offer a simple and non-invasive authentication approach for daily use. However, PPG authentication faces reliability issues due to motion artifacts from physical activity and physiological variability over time. To address these challenges, we propose MTL-RAPID, an efficient and reliable PPG authentication model, that employs a multitask joint training strategy, simultaneously assessing signal quality and verifying user identity. The joint optimization of these two tasks in MTL-RAPID results in a structure that outperforms models trained on individual tasks separately, achieving stronger performance with fewer parameters. In our comprehensive user studies regarding motion artifacts (N = 30), time variations (N = 32), and user preferences (N = 16), MTL-RAPID achieves a best AUC of 99.2\% and an EER of 3.5\%, outperforming existing baselines. We opensource our PPG authentication dataset along with the MTL-RAPID model to facilitate future research on GitHub.

Figures (13)

• Figure 1: Optical Principle. PPG signals contain two key components: non-physiological variations $m(t)$, influenced by external factors like lighting, body movements, and jitter, and physiological variations $p(t)$, driven by heart activity, vascular structure, and skin anatomy.
• Figure 2: MTL-RAPID architecture. We propose a multi-task RAPID model and switch training procedure to accomplish waveform quality assessment and identity verification tasks simultaneously.
• Figure 3: Procedure for data collection. ANT motion dataset fetures 10 activities (ranging from 60 to 300 seconds) with the re-worn process.
• Figure 4: Devices in the user study.
• Figure 5: The results of cross-activity experiments on ANT_Maxim dataset. MTL-RAPID got the best AUC of 99.2% and an EER of 3.5%, outperforming all baselines. The x-axis represents the pass rate after filtering, and the y-axis displays the AUC and EER of the identity verification task.