面向钓鱼检测、自洽性、忠实性与可解释性的大语言模型评估
Evaluating Large Language Models for Phishing Detection, Self-Consistency, Faithfulness, and Explainability
AI总结：
本研究针对钓鱼检测中LLM预测准确性与解释可信自洽性的问题，微调BERT、Llama、Wizard等模型，用CC SHAP度量评估，发现Llama解释对齐度高但准确率低，Wizard则相反。
摘要：
钓鱼攻击仍是最普遍、最持久的网络安全威胁之一，攻击者不断演化并强化策略以规避通用检测系统。尽管人工智能和机器学习已取得显著进展，但忠实复现支撑钓鱼判断的可解释推理、分类及可解释性仍颇具挑战。随着自然语言处理领域的最新进展，大语言模型（LLM）为改进特定领域的钓鱼分类任务展现出了可观的方向与潜力。然而，提升分类模型的可靠性与鲁棒性，不仅需要LLM给出准确预测，还需要与预测相一致的、可信的解释。因此，一个关键问题依然存在：LLM能否不仅准确分类钓鱼邮件，还能生成与其预测可靠对齐、且内部自洽的解释？为解答这些问题，我们微调了基于transformer的模型，包括BERT、Llama系列模型和Wizard，以提升领域相关性，使其更适配钓鱼场景的特定区分需求，所采用的技术包括二元序列分类、对比学习（CL）和直接偏好优化（DPO）。为此，我们应用基于SHAPley值的一致性度量（CC SHAP）评估了这些模型在钓鱼分类与可解释性方面的表现，该度量通过计算预测与解释的token对齐度，测试模型的内部忠实性与一致性，并揭示其预测与推理背后的逻辑。总体而言，我们的研究结果表明：Llama模型尽管决策准确率欠佳，但预测与解释的token对齐度更强，CC SHAP得分更高；而Wizard取得了更优的预测准确率，但CC SHAP得分更低。
英文摘要：
Phishing attacks remain one of the most prevalent and persistent cybersecurity threat with attackers continuously evolving and intensifying tactics to evade the general detection system. Despite significant advances in artificial intelligence and machine learning, faithfully reproducing the interpretable reasoning with classification and explainability that underpin phishing judgments remains challenging. Due to recent advancement in Natural Language Processing, Large Language Models (LLMs) show a promising direction and potential for improving domain specific phishing classification tasks. However, enhancing the reliability and robustness of classification models requires not only accurate predictions from LLMs but also consistent and trustworthy explanations aligning with those predictions. Therefore, a key question remains: can LLMs not only classify phishing emails accurately but also generate explanations that are reliably aligned with their predictions and internally self-consistent? To answer these questions, we have fine-tuned transformer based models, including BERT, Llama models, and Wizard, to improve domain relevance and make them more tailored to phishing specific distinctions, using Binary Sequence Classification, Contrastive Learning (CL) and Direct Preference Optimization (DPO). To that end, we examined their performance in phishing classification and explainability by applying the ConsistenCy measure based on SHAPley values (CC SHAP), which measures prediction explanation token alignment to test the model's internal faithfulness and consistency and uncover the rationale behind its predictions and reasoning. Overall, our findings show that Llama models exhibit stronger prediction explanation token alignment with higher CC SHAP scores despite lacking reliable decision making accuracy, whereas Wizard achieves better prediction accuracy but lower CC SHAP scores.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
