Hand authentication on multi-touch tablets

Abstract

We propose a biometric authentication scheme suitable for multi-touch devices such as tablet computers. Our scheme is based on hand geometry. It improves on prior work by introducing a dynamic element, where movement challenges are issued based on static hand geometry data. Specifically, we demonstrate a set of multi-touch interactions that can capture hand geometry information of users. For each of the interactions, we extract different but complimentary hand geometric information from the user. Our approach has several advantages over traditional text password and other biometric authentications. First, unlike other recognition based authentication schemes, a user is only expected to interact with the multi-touch surface according to the challenges she is posed. In other words, she does not have to memorize any type of credential. In addition, the system provides security against replay attacks--which is a drawback associated with many authentication schemes, such as traditional biometric system based on recognition of face, iris or fingerprint. Last but not least, our approach works on current tablet computers without any needs for updates of hardware, firmware or drivers -- it can be carried out by an application. We demonstrate experimentally a configuration using 14 consecutive challenges on iPad2 tablet (taking approximately 3.32 minutes for novice users to respond to), wherein the user is authenticated with almost 97% accuracy.