Context-based Clustering to Mitigate Phishing Attacks

Phishing is by far the most common and disruptive type of cyber-attack faced by most organizations. Phishing messages may share common attributes such as the same or similar subject lines, the same sending infrastructure, similar URLs with certain parts slightly varied, and so on. Attackers use such strategies to evade sophisticated email filters, increasing the difficulty for computing support teams to identify and block all incoming emails during a phishing attack. Limited work has been done on grouping human-reported phishing emails, based on the underlying scam, to help the computing support teams better defend organizations from phishing attacks. In this paper, we explore the feasibility of using unsupervised clustering techniques to group emails into scams that could ideally be addressed together. We use a combination of contextual and semantic features extracted from emails and perform a comparative study on three clustering algorithms with varying feature sets. We use a range of internal and external validation methods to evaluate the clustering results on real-world email datasets. Our results show that unsupervised clustering is a promising approach for scam identification and grouping, and analyzing reported phishing emails is an effective way of mitigating phishing attacks and utilizing the human perspective.