Nowadays the use of both Web Pages and Applications is constantly growing and browser-based malicious infections are extensively widespread, getting the end users' devices more and more vulnerable. The main purpose of this master thesis research is to implement an almost completely automatic methodology to recognize JavaScript malware families and to assist analyst on knowing about this cyber-security domain. The model is tested on the 20% of the same malware family files with a k-fold, other malicious but different type file and a set of good samples. In this phase it expects the model recognizes the specific malware type from the rest, classifying good samples and other malware in the same way as outliers.