This system demonstrates enhanced detection accuracy against phishing emails using machine learning, highlighting resilience against LLM-generated content.
Key Points
Our approach achieves a detection accuracy of 94.26%, successfully identifying both conventional and LLM-generated phishing emails.
Text preprocessing methods, including spelling correction and word splitting, enhance the model’s ability to detect adversarial modifications.
Evaluation against adversarial phishing samples shows the model's robustness against evolving AI threats and detection challenges.
Integration of machine learning algorithms and NLP techniques delivers improved performance over traditional phishing detection systems.