The increasing technological advancement and use of smartphones allows users to store a wide range of sensitive information, such as bank details, documents, social media data, passwords, contacts, among others. However, the loss or theft of these devices, along with vulnerabilities like password attacks, can seriously compromise the security of this data. Although traditional security mechanisms such as passwords and pattern locks have been widely used, their effectiveness has been questioned due to how easy they are to copy, forget, or share. Furthermore, they do not guarantee the link between an operation and the individual who carries it out. In response, many modern devices have adopted biometric technologies such as facial and fingerprint recognition, but their availability and cost can be limiting as their implementation requires special internal hardware which incurs more cost. Currently, machine learning has proven to be very effective in solving various problems in different areas. In this context, as another security alternative for these devices, we propose a low-cost and nonintrusive method using machine learning to identify users based on the dynamics of text typing on smartphones. To this end, we developed an Android application to collect typing data from 25 users, including the duration of each key pressed and the latencies between consecutive keys. We train and test models using algorithms such as Random Forest, XGBoost and LightGBM. The results show that the XGBoost model achieved the best performance, with an accuracy of 86.47% on the test set and 90.5% in a real environment using an API linked to the Android application.

Typing dynamics, Biometrics, Smartphones, Machin Learning