The Application of Machine Learning Algorithms in Classification of Malicious Websites

This chapter compares three different machine learning techniques, i.e. Gaussian process classification, decision tree classification and support vector classification, based on their ability to learn and detect the attributes of a malicious website. The data used have all been sourced from HTTP headers, WHOIS lookups and DNS records. As a result, this does not require parsing of the website content. The data are first subjected to multiple steps of pre-processing including data formatting, missing value replacement, scaling and principal component analysis