LLM-Based Framework for Email Classification and Phishing Detection
Nuno Costa, José Cecílio, Ruben Salgueiro, Mauricio Rosa, Dulce Domingos · ACM SIGAda Ada Letters · 2025
Phishing and spam attacks continue to pose significant security risks due to their evolving nature, with traditional email classification methods often unable to effectively address these challenges. This paper proposes a framework that uses Large Language Models (LLMs) for enhanced email categorization and phishing detection. Through an inbox plugin, a user reports an email, which is then stored in a database, cleaned, and thoroughly analyzed using a series of specific prompts determining a classification of "normal", "unwanted", or "malicious" email.