EVALUATING MODEL TRANSPARENCY IN PHISHING DETECTION USING LIME EXPLAINABILITY FRAMEWORKS
Keywords:
phishing, Explainable AI, LIME, model transparency, cybersecurity, AI frameworks, URLs, Security Operation CentersAbstract
Blackbox machine learning models are widely deployed in cybersecurity for mostly automating threats but due to their lack of interpretability it has become tough for security analysts to determine why specific web link or email content is flagged as a phishing. In modern enterprise environments, security operations centers (SOCs) and incident response teams usually handle the massive volumes of incoming digital traffic daily but they mostly rely on complex classification algorithms and pipelines to filter out malicious URLS and misleading payloads. For example, when transformer models flag specific email as malicious with high confidence , LIME reveals this decision was made due to entirely one typo. This research investigates the practical implementation of the LIME (Local Interpretable Model-agnostic Explanations) framework to suggest visualizing complex decision making ways of modern phishing detection. Along with that LIME highlights the exact sub-components – such as unusual strings, suspicious domains or parameters that trigger specific model results. After that security analysts can observe specific predictions closely, helping them to differentiate between legitimate and malicious content.
References
1.Abdelhamid, N., Ayesh, A., & Thabtah, F. (2014). Phishing detection based on associative classification data mining. Expert Systems with Applications, 41(13), 5948–5959. https://doi.org/10.1016/j.eswa.2014.03.019
2. Aljofey, A., Jiang, Q., Qu, Q., Huang, M., & Niyigena, J. P. (2022). An effective phishing detection model based on character-level convolutional neural network from URL. Electronics, 11(1), 60. https://doi.org/10.3390/electronics11010060
3. Anti-Phishing Working Group. (2024a). Phishing activity trends report, Q4 2024. APWG. https://apwg.org/trendsreports/
4. Bahnsen, A. C., Torroledo, I., Camacho, L. D., & Villegas, S. (2017). DeepPhish: Simulating malicious AI. Proceedings of the APWG Symposium on Electronic Crime Research (eCrime), 1–8. https://doi.org/10.1109/ECRIME.2017.7945461
5. Basnet, R. B., Mukkamala, S., & Sung, A. H. (2014). Detection of phishing attacks: A machine learning approach. Soft Computing Applications in Industry. https://doi.org/10.1007/978-3-540-77465-5_19
6. Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324
7. Canadian Institute for Cybersecurity. (2025). CIC-Trap4Phish dataset [Dataset]. University of New Brunswick. https://www.unb.ca/cic/datasets/
8. Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. Proceedings of NAACL-HLT 2019, 4171–4186.