DescriptionNowadays user authentication has been deployed at almost every corner of our daily lives. In this thesis, we propose a vibration-based user authentication system, VibAuth, which exploits the unique human behavior and physiological features embedded in the user’s finger-input vibrations. Particularly, our system actively transmits chirp vibration signals via a regular solid surface (e.g., table) and senses the vibrations on the surface at the same time through piezoelectric bimorph sensors. The user’s finger-input behavior (e.g., pressing position, touching force, contact area, and biophysical structure of the finger) makes distinct changes to the received signals in the frequency domain. VibAuth then extracts unique vibration features and uses a deep learning-based approach to differentiate between distinct people. Comparing to the traditional authentication approaches (e.g., PIN password and lock pattern), VibAuth enables user authentication on any solid surface via a novel touch sensing technique. The proposed two-factor authentication system greatly improves the security levels by combining the credentials with unique human finger-input physiological and behavior features. Extensive experiments demonstrate that VibAuth can achieve an average true positive rate of over 97% and a low false positive rate of 2%. It is also resilient to various attack scenarios including credential-targeted attacks, sniffing attacks, and side-channel attacks.