Explaining Real-Time Phishing and Explainable AI: A Transformer-Based Approach

Sonawane, Kunal Ramesh (2025) Explaining Real-Time Phishing and Explainable AI: A Transformer-Based Approach. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (605kB) | Preview
Preview	PDF (Configuration Manual) Download (423kB) | Preview

Abstract

Phishing is a growing cyber threat and the Anti-Phishing Working Group (APWG) recorded the highest number of 1,003,924 attacks in Q1 2025. The conventional static detection systems are easily overcome by advanced evasion methods, such as the rapid infrastructure changes and zero-day threats. This requirement makes sophisticated detection and the analysis of the URL string in real time. The current models with high-accuracy are still susceptible to unclear techniques such as Browser in the Browser and malicious embedded links. Most importantly, deployable systems should be transparent, and this necessitates integrating explainable details.

In order to evaluate, several models, such as the Logistic Regression, Random Forest, the XGBoost, and the RoBERTa-based embedding pipeline, were trained. The experiments prove that all models that are augmented with the transformer are much better than the traditional lexical models. XGBoost provided good performance with accuracy 95.49% and the highest inference speed of 0.14 ms per URL, which is highly suitable regarding the implementation in real-time. The SHAP integration offered understandable descriptions of model decisions with malicious subdomains, obfuscated paths, and encoded parameters being the most important. These findings confirm that the proposed system is accurate, low-latency, and interpretable, which are the main quality criteria of real-life phishing URL detection.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Sahni, Vikas UNSPECIFIED
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > QA Mathematics > Computer software > Computer Security T Technology > T Technology (General) > Information Technology > Computer software > Computer Security
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	09 Sep 2026 09:53
Last Modified:	09 Sep 2026 09:53
URI:	https://norma.ncirl.ie/id/eprint/9916

Actions (login required)

View Item