A Hybrid Phishing Email Detection Framework Leveraging Transformer-Based Embeddings, NIST PhishScale, and Ensemble Learning
Open Access DepositedA hybrid phishing email detection framework will leverage multiple methods to improve accuracy and vigor in identifying malicious emails. Machine learning and systems that are rule-based are some of these techniques. This framework will detect suspicious URLS and spoofing of domains by employing subtle language changes. Phishing efforts are captured by using OpenAI’s text-embedding-3-large, NIST PhishScale cues and structured email properties. Random Forest, XGBoost, and Support Vector Machine classifiers in this study are combined to demonstrate the effectiveness within a soft voting ensemble model to improve phishing detection behavior. There is a 99.74% accuracy rate by incorporating embedding that is generated by OpenAI’s text-embedding-3-large which highlights promise of this approach for phishing detection. The findings further emphasize the limitations of individual machine learning classifiers and demonstrate the robustness of ensemble methods in addressing different phishing tactics. In addition, including NIST PhishScale cues enhances detection by capturing psychological and linguistic deception patterns often overlooked by conventional models.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.