Biometric Signature Authentication with Low Cost Embedded Stylus.

This paper presents an affordable stylus device with embedded inertial sensing for measuring dynamic kinematic data from handwritten signatures for the purposes of user identification. A set of spatiotemporal features are proposed for use in a simple multilayer perceptron classifier, and a brief user study is conducted for evaluation with promising results. In general, user authentication is a key component of securing digital information in cyber-physical systems. Current methods span alphanumeric passwords, multi-factor authentication, and biometric techniques, with each providing trade-offs between convenience, flexibility and security. This work presents a device that marries a kinematic trajectory unique to each person (handwritten signature) with digital authentication via a stylus type device. Handwritten signatures are ubiquitous for authenticating paperwork, credit card transactions, check deposits and ballot boxes to name a few. Oftentimes, handwritten signatures are executed and treated perfunctorily as a matter of routine with no genuine intention towards security or authentication. When authentication is requested of handwritten signatures, most often the only recourse is expert visual examination of the written pen strokes. While some methods have investigated the use of measuring temporal, inertial data from handwritten signatures as a mode of authentication, these were executed with either expensive haptic robotic devices or prototype, externally mounted sensors. This work enables dynamic inertial authentication methods of handwritten signatures in a low cost, seamless embedded stylus device.