Numeric-Passcode Keystroke Biometric Studies on Smartphones

A keystroke biometric classification system traditionally used on data captured from physical keyboards associated with laptops and personal computers, was extended to evaluate biometric data extracted from mobile devices. The results were compared to the results of similar studies utilizing the same data inputs. Additional results were extracted; including results tied to features native to mobile devices, as well as results of the combined keystroke and mobile (touchscreen) feature file. The best results observed in this study were associated with the touchscreen biometric features, which achieved a top Equal Error Rate (EER) of 4.9%. As a result of the combined feature results being uncharacteristically higher than the touchscreen features alone, further study will include identifying other distance metrics that would better suit the feature data associated with this study. Keywords—biometrics; pattern recognition; machine learning; keystroke dynamics; mobile devices; user authentication