High precision handedness detection based on short input keystroke dynamics
Avar Pentel · 2017
Handedness is one of the soft biometrie attributes that can be employed for profiling anonymous authors. In this study, we present handedness detection by typing differences. We collected free form keystroke data from 504 people, 101 reported left hand as their writing hand and 403 were right handed. Using machine learning we discriminated left and right handed people over 99.5% accuracy with balanced dataset. Finally, we reduced feature set to a single feature which still yield to 98% classification accuracy.