Continuous authentication (CA) is the process to verify the user’s identity regularly without their active participation. CA is becoming increasingly important in the mobile environment in which traditional one-time authentication methods are susceptible to attacks, and devices can be subject to loss or theft. The existing literature reports CA approaches using various input data from typing events, sensors, gestures, or other user interactions. However, there is significant diversity in the methodology and systems used, to the point that studies differ significantly in the features used, data acquisition, extraction, training, and evaluation. It is, therefore, difficult to establish a reliable basis to compare CA methods. In this study, keystroke mechanics of the public HMOG dataset were used to train seven different machine learning classifiers, including ensemble methods (RFC, ETC, and GBC), instance-based (k-NN), hyperplane optimization (SVM), decision trees (CART), and probabilistic methods (naïve Bayes). The results show that a small number of key events and measurements can be used to return predictions of user identity. Ensemble algorithms outperform others regarding the CA mobile keystroke classification problem, with GBC returning the best statistical results.