Full Program

LLM-Powered Intent-Based Categorization of Phishing Emails

Summary:

This paper investigates the potential of using large language
models (LLMs) to detect these emails by focusing on their intent.
In addition to binary classification of phishing emails, the paper
explores LLM to classify emails into distinct intent categories, in
which we develop a taxonomy, and extract actionable cyber threat
intelligence. To support our work, we develop a custom dataset
containing a mix of legitimate and phishing emails curated from
publicly available datasets. Our results show that existing LLMs
can detect and categorize phishing emails, highlighting their
potential for this domain.

Author(s):

Even Eilertsen    
University of Oslo
Norway

PhD Researcher at University of Oslo

Vasileios Mavroeidis    
University of Oslo
Norway

Gudmund Grov    
Norwegian Defence Research Establishmend (FFI)
Norway