Tools
Tools
Bhavana, B. S., Srivastava, Shatakshi, Ghosh, Debjani, Kumar, Vimal and Gupta, Punit (2025) Context-Aware Phishing Email Detection Using Hybrid Machine Learning and Explainable AI. In: 2025 Seventeenth International Conference on Contemporary Computing (IC3). IEEE, Noida, India, pp. 1-6. ISBN 979-833155901-4
Full text not available from this repository.Abstract
Phishing emails remain a major cybersecurity threat, often bypassing traditional spam filters due to their deceptive and evolving nature, such as AI-generated content, personalized social engineering, URL obfuscation, and advanced credential harvesting tactics. This study proposes a hybrid machine learning framework for effective phishing detection, combining classical classifiers like Random Forest, Naïve Bayes and XGBoost, deep learning models like CNN, LSTM and CNN-LSTM as well as the transformer-based DistilBERT model. A unified and balanced corpus was created by merging three publicly available datasets. It was then preprocessed, and applied TF-IDF vectorization and tokenization for different models. SMOTE was used to address class imbalance. Among all models, DistilBERT achieved the highest accuracy of 98.2%, benefiting from its strong contextual understanding. To enhance trust and transparency, Explainable AI tools like LIME, SHAP, and BERTViz were used to interpret model decisions. The proposed pipeline offers a scalable and interpretable solution for real-time phishing email detection, with strong potential for future applications in enterprise security.
Actions (login required)
 |
View Item |