← Back to VOLUME 14, ISSUE 11, NOVEMBER 2026
This work is licensed under a Creative Commons Attribution 4.0 International License.
Fraudulent Website Detection Using XGBoost and Lexical URL Features with Explainable Risk-Based Alert Presentation
Tarsha P, S. Anitha Jebamani, Snifa Rose Vilson, Neha Raj, Vandhana Prabhakar
π 1 viewπ₯ 1 download
Abstract: Phishing websites imitate legitimate services to steal credentials, and blacklist-based defences are slow to cover newly created domains. This paper presents a fraudulent website detection system that classifies a URL as legitimate or potentially phishing from 19 lexical features alone, so that no web page has to be downloaded. An Extreme Gradient Boosting (XGBoost) classifier is trained on the PhiUSIIL Phishing URL dataset (235,795 records); each prediction is converted into a phishing probability, a risk score and level, and risk-attribution cards, and the pipeline is served through a Streamlit application for single and batch analysis. On a held-out test set of 12,000 unseen URLs the model reaches 99.60% accuracy, 99.44% precision, 99.87% recall (legitimate class as positive), a 99.65% F1-score and a ROC-AUC of 0.9984, with 39 phishing URLs missed and 9 legitimate URLs flagged, and it trains in 0.49 s on a CPU. Three features carry almost all of the importance and no independent data set was used, so the result is specific to this benchmark and requires external validation.
Keywords: Phishing detection, fraudulent websites, URL classification, lexical features, XGBoost, gradient boosting, explainable machine learning, risk scoring, PhiUSIIL
Keywords: Phishing detection, fraudulent websites, URL classification, lexical features, XGBoost, gradient boosting, explainable machine learning, risk scoring, PhiUSIIL
How to Cite:
[1] Tarsha P, S. Anitha Jebamani, Snifa Rose Vilson, Neha Raj, Vandhana Prabhakar, βFraudulent Website Detection Using XGBoost and Lexical URL Features with Explainable Risk-Based Alert Presentation,β International Journal of Innovative Research in Electrical, Electronics, Instrumentation and Control Engineering (IJIREEICE), DOI: 10.17148/IJIREEICE.2026.141104
