Evaluating Large Language Models for Phishing Detection, Self-Consistency, Faithfulness, and Explainability
This work addresses the need for more reliable and interpretable phishing detection systems for cybersecurity applications, though it appears to be an incremental improvement focusing on specific model comparisons.
The paper tackles the problem of improving phishing email detection by evaluating whether large language models (LLMs) can accurately classify phishing emails while generating explanations that are self-consistent and faithful to their predictions, using fine-tuned models like BERT, Llama, and Wizard with methods such as contrastive learning and direct preference optimization. The results show that Llama models achieve higher explanation consistency scores (CC SHAP) despite lower accuracy, while Wizard has better accuracy but lower consistency scores.
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.