User Identification through Hidden Markov Model-based Touch Keystroke Dynamics
Umberto Bilotti, Carmen Bisogni, Aniello Castiglione, Michele Nappi, Chiara Pero · 2023
The biometric feature for user identification based on user typing patterns on touchscreen devices, modeled by Hidden Markov Models (HMMs), has shown to be highly effective and accurate. HMMs are well-suited for this purpose due to their stochastic nature, which allows for revealing pattern distributions and predicting possible keystroke sequences. To assess the performance of our proposed approach, we conducted experiments on two publicly available datasets. The results showed that our method achieved high accuracy rates and effectiveness in identifying users based on their typing patterns. In fact, the accuracy rates achieved by HMMs were significantly higher than those obtained by well-known machine learning algorithms, reaching up to 94% and 98%. Overall, our results demonstrate that the proposed biometric feature, based on user typing patterns on touchscreen devices and modeled by HMMs, is a powerful and reliable tool for user identification. This approach could be useful in a variety of applications, such as authentication and access control systems, where accurate user identification is critical.