EUVIMEDEUVIMED entdecken
Crossref

Analyzing Robustness and Explainability in URL Fraud Detection Using a Multi-Model Ensemble and RAG-Augmented Transformer Framework

B.Anitha, V.Munni

Research Digest on Engineering Management and Social Innovations · 31.03.2026

DOI: 10.46647/icetetas041 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Research Digest on Engineering Management and Social Innovations, online erschienen am 31.03.2026. Verfasst von B.Anitha und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

With the rapid advancement in information technology, many aspects of our lives are being affected, especially regarding the way we interact with online systems. In this regard, online systems such as internet banking services, online shopping sites, cloud computing services, as well as social media applications, are essential aspects of our day-to-day activities. Although online systems have greatly contributed to our convenience and efficiency in conducting online operations, they are associated with a number of risks regarding cyber threats. In the online world, the most risky situation we encounter includes online phishing, where attackers try to trick internet users into sending them information through sending malicious URLs in the form of legitimate online sites. These online attackers cleverly create the URLs in a way that they are legitimate enough so that they can trick internet users into sending them information such as usernames, passwords, as well as financial information. In the past, online phishing sites were detected through the application of a blacklist, where a number of identified online phishing sites are stored in a database to filter out online phishing sites in the future. Although this approach has greatly contributed to the detection of online phishing sites, this approach has been identified as being less efficient in detecting new online phishing sites, as online attackers cleverly change the domain names as well as URL formats of online phishing sites in a way that they are not detected by this approach. Therefore, advanced approaches in online phishing site detection are required in order to effectively handle this situation of online phishing sites. In this study, a hybrid approach has been proposed for the detection of online phishing sites by using a combination of several machine learning algorithms and several deep learning algorithms. In this approach, the decision tree model XGBoost has been implemented for the analysis of URL characteristics, and the character-level CNN model has been implemented for the analysis of hidden characteristics in URL strings using the transformer encoder model. In addition to this, the proposed framework has also taken into consideration the explanation factor. Artificial intelligence systems are considered to be a black box model, and it is very difficult to understand the prediction mechanism. In this direction, the proposed framework has included the Retrieval-Augmented Generation module through the Large Language Model.

Zitieren

APA

B.Anitha, & V.Munni (2026). Analyzing Robustness and Explainability in URL Fraud Detection Using a Multi-Model Ensemble and RAG-Augmented Transformer Framework. Research Digest on Engineering Management and Social Innovations. https://doi.org/10.46647/icetetas041

Vancouver

B.Anitha, V.Munni. Analyzing Robustness and Explainability in URL Fraud Detection Using a Multi-Model Ensemble and RAG-Augmented Transformer Framework. Research Digest on Engineering Management and Social Innovations. 2026. doi:10.46647/icetetas041

BibTeX
@article{BAnitha2026,
  author = {B.Anitha and V.Munni},
  title = {Analyzing Robustness and Explainability in URL Fraud Detection Using a Multi-Model Ensemble and RAG-Augmented Transformer Framework},
  journal = {Research Digest on Engineering Management and Social Innovations},
  year = {2026},
  doi = {10.46647/icetetas041}
}
Kostenlos starten →