Nowadays, due to technology enhancement faults are increasingly compromising all kinds of computing machines, from servers to embedded systems. Many methods have been proposed in these years, however, machine learning brought many possibilities to empower faults detection exploiting hardware metrics inspection, and it is now possible to explore the opportunity of avoiding the use of heavy software techniques or product-specific errors reporting mechanisms. In this thesis, data analysis will be performed on several datasets collected by many simulated runs, with and without faults injection. The final goal of this work is to find the best machine learning model to cope with the task and ultimately to build an initial implementation of a monitoring tool, able to detect faults process-wise using pre-trained models on hardware metrics extracted by a kernel module..