The aim in this research is to explore the possibility of Identifying and Verifying computer users by extracting and analyzing their keyboard typing patterns. We study the performances of four novel architectures based on Naı̈ve Bayes and the L2 Norm metric, both modeled to solve identification and verification tasks. In order to do this, we explore a variety of data collection techniques and show how a mixture of the considered ones leads to optimized results. We then evaluate all the resulting models by computing the most commonly used metrics for each task, and comparing them with those achieved by Giot, El-Abed and Rosenberger.