Method of learning malware behavior scripts by sequential pattern mining

Abstract

Scripts are the knowledge representation model. To our knowledge, there were no machine learning methods for it. In this work we propose a method of discovering script goals and putting them into an order. It is based on sequential pattern mining and regular expressions. This method has been validated by experiments set on malware behavior data. The results show that the discovered goals and their order correspond with expected malware behavior.