A Generalized Authentication Scheme for Mobile Phones Using Gait Signals

Despite the reliability of authentication schemes using tokens or biometric modalities, their requirement of explicit gestures makes them less usable. On the other hand, the study on gait signals which are potential reliable for effective implicit authentication have been raised recently. Having said that, all the existing solutions fail to be applicable in reality since they rely on having sensors fixed to a specific position and orientation. In order to handle the instability of sensor’s orientation, a flexible approach taking advantages of available sensors on mobile devices is our main contribution in this work. Utilizing both statistical and supervised learning, we conduct experiments on the signal captured in different positions: front pocket and waist. In particular, adopting PCA+SVM brings about impressive results on signals in front pocket with an equal error rate of 2.45 % and accuracy rate of 99.14 % in regard to the verification and identification process, respectively. The proposed method outperformed other state-of-the-art studies.