Characterizing Phishing attacks using Natural Language Processing

Currently, there are many ways that phishing attacks placed people and businesses at risk. In the economic causes losses the money; in the social aspect, the perception of trust in users decreases; and on the psychological level, fear can avoid the use of digital tools and resources. This study aims to increase efficiency in detecting Phishing attacks using Natural Language Processing (NLP) to explore the mental model people use to detect whether an email is legitimate or not. Specifically, it is based on feedback vectorization and the movement of the mouse, which was obtained when the participants interacted in a test to detect phishing. The results obtained allow us to identify that people based their decision on the URL analysis in the mental model of legitimation and phishing decision. However, in the phishing model, the number of characteristics in each indicator would be more diverse and broader, which produces new challenges and future directions in this solution.