Keystroke dynamics, accelerometer and gyroscope data based behavioral biometric authentication on mobile devices

In mobile devices which we use frequently and can handle many high-importance works, fingerprint or face recognition systems are widely used as biometric authentication methods in directly use or as additional factors with the password authentication methods. Although these authentication methods appear to be reliable, biometric data that provides authentication cannot be hidden. Physical biometric data such as fingerprints, sounds and facials are permanently left in places or cannot be hidden from others. In other words, authentication is performed with data that is seen, heard, accessible by everyone else and theese are unchangeable. It is not known whether verification systems can be easily circumvented by stealing or copying this data in the future. Behavioral biometric authentication methods can be used in place of or optionally in addition to these methods, can be performed with certain success percentages by means of background behavior analyzes without the user having to deal with an additional verification step. In this study disscussed a biometric authentication solution named 'keystroke dynamics and device position data during keying in three-dimensional space obtained by accelerometer and gyroscope sensors on mobile devices' which is much more difficult to steal and imitate and which is more efficient in terms of user experience because it works transparently in the background. The analyzes were performed using different machine learning algorithms and different data set combinations. According to the average results, the best equal error rate of 1.30% was obtained with Random Forest Classifier.