WebGuard AI-Powered Cyber Threat Detector Using BERT And AutoEncoder

Read full text Summarise with VIRAT
Authors: Ravi M, Dr. A. Obulesu, A. Sathvik Samuel, K. Pravallika, S. Pavan, L. Anjaneyulu
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In the fast development of web applications, cyber attacks like phishing and poisonous web requests have become more advanced. There are old signature based detection methods which find it difficult to recognize new and zero day. attacks. In this paper, a cyber threat detection framework based on the use of AI and combining AutoEncoder-based anomaly detection with semantic phishing based on BERT is introduced. detection. The AutoEncoder model uses normal network behavior as a form of learning and detects the abnormal patterns, but it does not need labeling data, whereas the BERT model carries out deep semantic analysis of URLs to distinguish between phishing attempts. The combination of the two methods allows the proposed system to be able to detect the known and those threats with high levels of robustness. Experimental analysis shows a high accuracy and low false positives and high generalization ability than conventional. techniques of machine learning.

Fields of Study & Subjects

Spam and Phishing Detection Network Security and Intrusion Detection Adversarial Robustness in Machine Learning

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
22 March 2026
Language
EN