COMPONENT FEATURES BASED ENHANCED PHISHING WEBSITE DETECTION SYSTEM USING EfficientNet, FH-BERT, AND SELU-CRNN METHODS

Mahmoud Murhej^*G Nallasivan^*

• Vel Tech Rangarajan Dr.Sagunthala R&D Institute of Science and Technology, Chennai, India

Phishing is a kind of cybercrime, which is induced by hackers to steal sensitive information of users. So, it becomes essential to detect phishing attacks on websites. Many prevailing works utilized the Uniform Resource Locator (URL) link and Document Object Model (DOM) tree structures for Phishing Website Detection (PWD). However, these existing works caused inaccurate detection results as the phishing website is an imitation of a legitimate website. Therefore, the proposed PWD system is developed by focusing on the important features and components of the websites to enhance the detection efficiency. The proposed work begins with the collection of URL links from the phishing website dataset. Then, the Hypertext Markup Language (HTML) formats are created for those URLs. Afterward, the DOM tree structure is constructed from the HTML format, and the components are extracted from it along with Natural Language Processing (NLP), credentials, URL, DOM tree similarity, and component features extractions. Meanwhile, the DOM-tree components are converted into score values by utilizing the Feature Hasher-Bidirectional Encoder Representations from Transformers (FH-BERT). Further, the component features and the score values of the DOM-tree components are concatenated by the feature fusion process. Thereafter, the significant features are selected using an Entropy-based Chameleon Swarm Algorithm (ECSA). Lastly, the phishing website is detected by employing the SELU-CRNN. The simulated results exhibited that the proposed technique improved the detection performance of PWD with higher accuracy (98.42%) and minimum training time (63003ms) than the prevailing methods.