Authors :
Padma Priya S.; Swetha R.; Thirishya M.
Volume/Issue :
ICMST-2025
Google Scholar :
https://tinyurl.com/26ss6fp9
Scribd :
https://tinyurl.com/2u8rzndy
DOI :
https://doi.org/10.38124/ijisrt/25nov754
Abstract :
Phishing remains one of the most widespread and evolving cyber threats, deceiving users through fake websites
that mimic legitimate platforms to steal sensitive information. Traditional detection methods such as blacklists, signature-
based filters, and manual verification fail to recognize newly emerging or obfuscated phishing sites. To overcome these
challenges, this paper presents HPD (Hybrid Phishing Detection), a real-time hybrid machine learning framework that
integrates heuristic analysis with advanced classification algorithms to improve accuracy and response time. The system
extracts and analyzes multiple feature sets including lexical features from URLs, host-based parameters from domain
registration data, and content-based attributes from webpage HTML structure and scripts. These features are processed
using a hybrid ensemble model combining Random Forest, Support Vector Machine (SVM), and Logistic Regression,
ensuring higher detection precision and reduced false positives. Experimental analysis using benchmark phishing datasets
demonstrates that HPD achieves more than 97% accuracy, outperforming traditional single-model systems. The lightweight
and scalable design allows deployment as a browser extension or through RESTful APIs for real-time threat detection. By
enabling adaptive learning and integration with cloud-based threat intelligence, HPD offers a proactive and reliable solution
for combating modern phishing attacks and enhancing web security.
Keywords :
Phishing Detection, Cybersecurity, Hybrid Machine Learning, Real-Time System, Ensemble Learning, URL Analysis, Website Security.
Phishing remains one of the most widespread and evolving cyber threats, deceiving users through fake websites
that mimic legitimate platforms to steal sensitive information. Traditional detection methods such as blacklists, signature-
based filters, and manual verification fail to recognize newly emerging or obfuscated phishing sites. To overcome these
challenges, this paper presents HPD (Hybrid Phishing Detection), a real-time hybrid machine learning framework that
integrates heuristic analysis with advanced classification algorithms to improve accuracy and response time. The system
extracts and analyzes multiple feature sets including lexical features from URLs, host-based parameters from domain
registration data, and content-based attributes from webpage HTML structure and scripts. These features are processed
using a hybrid ensemble model combining Random Forest, Support Vector Machine (SVM), and Logistic Regression,
ensuring higher detection precision and reduced false positives. Experimental analysis using benchmark phishing datasets
demonstrates that HPD achieves more than 97% accuracy, outperforming traditional single-model systems. The lightweight
and scalable design allows deployment as a browser extension or through RESTful APIs for real-time threat detection. By
enabling adaptive learning and integration with cloud-based threat intelligence, HPD offers a proactive and reliable solution
for combating modern phishing attacks and enhancing web security.
Keywords :
Phishing Detection, Cybersecurity, Hybrid Machine Learning, Real-Time System, Ensemble Learning, URL Analysis, Website Security.
