Institutional Repository of Vilnius University
Didžiųjų kalbos modelių ir specializuotų neuroninių tinklų taikymas URL phishing atpažinimui /
Abstract
dc:descriptionThis bachelor’s thesis investigates the problem of phishing and its detection using modern machine learning algorithms. Since traditional security methods, such as blacklisting, oſten fail to identify newly created threats in realtime, the study focuses on the effectiveness of specialized neural networks and ensemble tree algorithms in identifying malicious URLs. The experimental study was conducted using the “PhiUSIIL” dataset, which, aſter pre processing, consisted of over 150000 unique records. Two models were implemented and compared: a MultiLayer Perceptron (MLP) and the “XGBoost” algorithm. The research results demonstrated that the “XGBoost” model achieved an exceptionally high classification accuracy of 96.28%. To ensure decision transparency, the LIME method was applied to interpret the “black box” decisions made by the models and to identify critical URL features. The findings of the study confirm that machine learning, combined with explainable AI tools, provides a reliable and transparent solution for combating phishing websites.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Pliuskus, Povilas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308338440&prefLang=en_US