Core Concepts
Large language models can effectively detect phishing emails with high accuracy and provide detailed explanations for their determinations.
Abstract
ChatSpamDetector introduces a system that leverages large language models (LLMs) to accurately detect phishing emails. By analyzing both email headers and bodies, the system can identify various deceptive strategies, including brand impersonation and social engineering tactics. The system provides detailed explanations for its determinations, assisting users in making informed decisions about suspicious emails. Evaluation experiments showed that ChatSpamDetector using GPT-4 achieved an accuracy of 99.70%, outperforming other models and baseline systems.
Stats
我々のシステムはGPT-4を使用して99.70%の検出精度を達成した。
ChatSpamDetectorは、詳細な説明を提供し、ユーザーが疑わしいメールについて情報を得るのに役立つ。
大規模言語モデルを活用することで、フィッシングメールを高い精度で検出できる。