Multiple Classifier Systems for Touch-based Continuous Mobile Device Authentication

In this research, we proposed a Dynamic Selection (DS) classification technique based on the concept of Multiple Classifier Systems to select the most optimum classifier(s) to perform user classification for touch-based continuous mobile device authentication. The technique generates a pool of accurate and diverse classifiers and select the most optimum ones from the pool to perform the classification task for each touch stroke. Our findings show that DS methods outperformed static classifiers with a relatively lower equal error rate and are more consistent than most static classifiers.