I Want to Know Your Hand

Abstract

We present HoldPass, the first system that can authenticate a user while they simply hold their phone. It uses the heart activity as biometric trait sensed via the hand vibrations in response to the cardiac cycle - a process known as ballistocardiography (BCG). While heart activity has been used for biometric authentication, sensing it through hand-based ballistocardiography (Hand-BCG) using standard sensors found on commodity mobile phones is an uncharted territory. Using a combination of in-depth qualitative analysis and large-scale quantitative analysis involving over 100 volunteers, we paint a detailed picture of opportunities and challenges. Authentication based on Hand-BCG is shown to be feasible but the signal is weak, uniquely prone to motion artifacts and does not land itself to the common approach of alignment-based authentication. HoldPass addresses these challenges by introducing a novel alignment-free authentication scheme that builds on asynchronous signal slicing and a data-driven algorithm for identifying a reduced set of features for characterizing a user. We implement HoldPass and evaluate it using a multi-modal approach: a large-case study involving 112 volunteers and targeted studies with a smaller set of volunteers over a period of several months. The data shows that HoldPass provides an authentication accuracy and user experience on par with or better than state-of-the-art systems with stronger requirements on hardware and/or user participation.