Improving the effectiveness of phishing detection Using lexical semantics; A machine-learning based approach

Many share the opinion that phishing emails should automatically be detected, such that these emails can be filtered out and do not end up in our inbox. However, a method that perfectly does this has not yet been found. Prior research describes several methods that attempt to identify phishing emails based on structural properties, but to our knowledge, a better alternative does not yet exist. In this thesis, we propose a method that allows us to filter out these emails based on lexical semantics. We make use of machine learning-based algorithms in combination with a technique that carries the name of word embeddings, to design a method that can be used in automatic email classification. By implementing this method, we can let our computers automatically filter emails by making a judgement based on the contents of the emails, just like how they are presented to us as human beings.