A Behavioral Biometrics User Authentication Study Using Motion Data from Android Smartphones
Javid Maghsoudi, Charles C. Tappert · 2016
This study examined the behavioral biometrics of smartphone motion to determine potential authentication accuracies on Android phones. The study used machine learning algorithms to analyze data from the accelerometer and gyroscope sensors. Android smartphone data were captured from sixty different individuals, resulting in a large collection of datasets for training and testing. The data were filtered by removing noise and segmented into motion intervals prior to feature extraction. The classification algorithms employed in the study were Multilayer Perception, k-Nearest Neighbor, Support Vector Machines, and Naïve Bayes. Authentication accuracies achieved ranged from 81% to over 97%.