User Authentication on Headset-Like Devices by Bioacoustic Signals

User Authentication on Headset-Like Devices by Bioacoustic Signals

연구 분야: Analysis

학회: International Workshop on Emerging Technologies for Authorization and Authentication

Reliable and transparent user authentication on sensor-rich devices, such as wearables, is a topical task today. Of special interest are methods based on bioacoustic signals processing, such as on-body active and passive acoustic sensing. These methods are attractive due to the relatively small aging effect of the captured bioacoustic signals and low battery consumption. This makes them promising candidates for on-device user authentication. Most recent researches in bioacoustic user authentication are aimed at active acoustic sensing. Practical usage of such methods requires adding of an additional electro acoustic transducer to wearables which is inappropriate for already commercialized devices. Methods of passive acoustic sensing allow for overcoming these limitations by capturing bioacoustic signals produced during person’s movements, for example wrist rotations. However, practical application of these methods requires usage of microphones with high sensitivity for capturing of weak acoustic signals. To overcome this limitation we suggest to perform passive sensing near the place with multiple joints, such as cervical vertebrae. The results of performance analysis proved effectiveness of proposed solutions, namely decreasing of False Rejection Rate (FRR) errors up to ten times in comparison with state-of-the-art solutions while preserving low False Acceptance Rate (FAR) values. Achieved values FAR \(=0.12\%\) and FAR \(=3.00\%\) for proposed solution conforms to the requirements for Secondary Tier of Android OS Tiered Authentication Model that makes the solution an attractive candidate for user authentication on the next-generation wearable devices.