Intellectual Property India Publishes Patent Application for Hybrid Machine Learning And Deep Learning System And Method For Url-Only Phishing Detection Using Probability-Level Fusion Of Handcrafted-Feature And Character-Level Models.

MUMBAI, India, Oct. 12 -- Intellectual Property India has published a patent application (202641119230 A) filed by Muddana Seshma; Marella Soma Sekhar; Bathula Jayanith; Ravada Midhun Kumar; Maddula Revanth Sri; Kothagundu Udaya Praneeth Gupta; and Jonnalagadda Surya Kiran on October 05, 2026, for Hybrid Machine Learning And Deep Learning System And Method For Url-Only Phishing Detection Using Probability-Level Fusion Of Handcrafted-Feature And Character-Level Models..

Inventors include Muddana Seshma; Marella Soma Sekhar; Bathula Jayanith; Ravada Midhun Kumar; Maddula Revanth Sri; Kothagundu Udaya Praneeth Gupta; and Jonnalagadda Surya Kiran.

The application for the patent was published on October 09, 2026, under issue no. 41/2026.

Abstract: 1 HYBRID MACHINE LEARNING AND DEEP LEARNING SYSTEM AND METHOD FOR URL-ONLY PHISHING DETECTION USING PROBABILITY- LEVEL FUSION OF HANDCRAFTED-FEATURE AND CHARACTER-LEVEL MODELS ABSTRACT A URL-only phishing detection system and method extract 29 lexical and structural features from a raw URL and classify them with an XGBoost model, while a Character-Level Convolutional Neural Network, trained independently, classifies the URL's character sequence. The two phishing probabilities are combined by equal-weight probability-level fusion and a 0.5 threshold gives the final label, without accessing webpage content. In an evaluated embodiment the hybrid achieved 99.75% accuracy, 100.00% precision, 99.51% recall, and a 0.9983 ROC-AUC on a held-out test set.

Disclaimer: Curated by HT Syndication.