User authentication method based on keystroke dynamics and mouse dynamics using HDA

Biometric authentication has advantages over traditional authentication based on passwords or pin number (PIN) in that it is based on the user's inherent characteristics which is not easily stolen or lost. Keystroke dynamics and mouse dynamics are biometrics that study the behavior patterns of human-computer interaction (HCI). Personal keystroke pattern and mouse-movement pattern are difficult to imitate and can, therefore, be used for personal identity authentication. Keystrokes and mouse movements can potentially authenticate users without affecting the use of computers and other devices to improve system security. In real environments, authentication methods that fuse keystroke dynamics and mouse dynamics are less accurate. In this paper, a new method of user authentication using complex real-environment HCI data is presented, which is called authentication adaptation network (AAN). In this method, heterogeneous domain adaptation (HDA) method is used for user authentication based on keystroke dynamics and mouse dynamics for the first time. All representative time windows and dimensionality reduction targets of keystroke dynamics features are compared to determine the parameters of AAN to ensure the robustness of the algorithm, and the effectiveness of the algorithm is demonstrated by validation experiments and comparison with the methods proposed in previous studies. Finally, experiments using the collected real-environment HCI dataset obtained 89.22% user authentication accuracy, which indicate that the proposed method achieves an encouraging performance.