VibID: User Identification through Bio-Vibrometry

User identification lays the foundations of the security protection and data privacy preservation of wearable devices. With proper user identification, wearable devices can adopt personalized settings for different users, and automatically label the corresponding data to protect user privacy. Moreover, it also can help prevent illegal user spoofing attacks, e.g., a malicious user asks someone else to wear his device so that he can use the collected data to claim medical benefits. Current user identification solutions proposed for wearable devices either rely on dedicated devices with high cost or require user intervention which is not convenient. In this work, we leverage the biovibrometry to enable a novel user identification solution, which only uses the existing low-cost sensors that are already available for most wearable devices. Our key idea is that, when human body is exposed to a vibration excitation, its response reflects the physical characteristics, i.e., the mass, stiffness and damping. Besides, due to users’ biological diversity, such characteristics of different users are quite distinctive. Therefore, we leverage the discrepancy in users’ vibration responses as an identifier. Based on this idea, we propose VibID, which only uses a lowcost vibration motor and accelerometer to simulate an unobtrusive vibration to users’ arms and capture the corresponding responses. By extracting the vibration patterns at different frequency, VibID builds an ensemble machine learning model to perform user identification. Extensive experiments are conducted to demonstrate that our system is robust to various confounding factors, including: arm position, muscle state, user mobility and wearing location. We also show that, in an uncontrolled scenario of 8 users, our system can still ensure an identification accuracy above 91%.