Handwriting Data Analysis from Crayonic KeyVault Smart Security Device

Abstract

In this article, we discuss handwriting as a behavioral biometric modality. The main contribution focuses on biometric data acquired from handwriting using the pioneering Crayonic KeyVault device, which has a built-in gyroscope, accelerometer, and pressure sensor. Thanks to these sensors, this device can record the movements of the hands while handwriting the digits based on the prompts of the device itself. This data in raw format was sampled and then normalized and parameterized. The time-sampled data were visualized and the intrapersonal similarity was observed from multiple typing attempts of the same participant, and the interpersonal variability when the same number was written from different participants. The data was then divided into a test and a training set. The training set was used to train the machine learning models and the accuracy was verified with the test set. The 91.6% accuracy of the person model identification on the test set was achieved.