PUBLICATIONS

ABSTRACT

This paper introduces a multi-modal authentication approach for payments using a face image and the low energy Bluetooth (BLE) signal of a user’s device. Devices of registered users transmit temporally changing one-time identifiers. During a payment request, the query face image is only matched with database entries corresponding to nearby users, thereby significantly reducing the complexity
of the task. For cases in which a user does not carry their device, the system includes a fallback mechanism to PIN based two-factor authentication. A classifier on depth data input is used to reduce vulnerability to presentation attacks.We conducted a user study of different payment methods and
demonstrated our system at a public event with 951 users.