User verification using safe handwritten passwords on smartphones

Abstract

This article focuses on the writer verification using safe handwritten passwords on smartphones. We extract and select 25 static and dynamic biometric features from a handwritten character password sequence on an android touch-screen device. For the writer verification we use the classification algorithms of WEKA framework. Our 32 test persons wrote generated safe passwords with a length of 8 characters. Each person wrote their password 12 times. The approach works with 384 training samples on a supervised system. The best result of 98.72% success rate for a correct classification, the proposal reached with the KStar and k- Nearest Neighbor classifier after ranking with Fisher Score feature selection. The best result of 10.42% false accepted rate is reached with KStar classifier.