StructureID: Implicit User Authentication by Sensing Structure-borne Sound on Smart Device

User authentication on mobile smart devices is essential for safeguarding personal privacy. However, existing methods based on facial recognition and fingerprint biometrics are susceptible to impersonation attacks, while those relying on PIN codes or unlock patterns are vulnerable to theft or observation. To address these issues, this paper proposes StructureID, an implicit authentication approach that leverages users' grip posture biometrics through active acoustic sensing on smart devices. By applying a series of signal processing techniques and deep learning models, StructureID effectively eliminates interference from direct airborne signals and multipath noise. It then extracts distinctive acoustic features from structure-borne signals within the device to enable secure user authentication. Experimental results demonstrate that StructureID achieves an authentication accuracy of nearly 90% and maintains a false acceptance rate below 8%.