Scholarlink

This paper introduces a hybrid approach to email phishing detection that integrates the META LLaMA Large Language Model (LLM) with a system of predefined trigger phrases, further enhanced through Bayes classifier fine-tuning. We evaluate several detection strategies on datasets containing both phishing and legitimate emails to identify the most effective solution. Our findings reveal that standalone LLMs are limited by lengthy preprocessing times and relatively low classification accuracy. In contrast, the proposed hybrid model significantly improves accuracy and provides robust categorization of phishing emails. These advantages make it a practical and scalable solution for deployment in real-world environments where precise classification is essential.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid Email Phishing Detection Using Large Language Models and Bayes Classifiers

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ciprian Chiosa,
  • Ciprian Pungila

摘要

This paper introduces a hybrid approach to email phishing detection that integrates the META LLaMA Large Language Model (LLM) with a system of predefined trigger phrases, further enhanced through Bayes classifier fine-tuning. We evaluate several detection strategies on datasets containing both phishing and legitimate emails to identify the most effective solution. Our findings reveal that standalone LLMs are limited by lengthy preprocessing times and relatively low classification accuracy. In contrast, the proposed hybrid model significantly improves accuracy and provides robust categorization of phishing emails. These advantages make it a practical and scalable solution for deployment in real-world environments where precise classification is essential.