MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202621084013 A) filed by Akhilesh Sharma; Mahi Tiwari; Poonam Kumari; Rupali Pawar; Tripty Kumari; Priyanshi; Shaziya Khan; Vivek Chouhan; Md. Masoom Ansari; and Shivani Banjare on July 08, 2026, for Ai-Based Intelligent Phishing Detection System Using Hybrid Nlp And Machine Learning Models.
Inventors include Akhilesh Sharma; Mahi Tiwari; Poonam Kumari; Rupali Pawar; Tripty Kumari; Priyanshi; Shaziya Khan; Vivek Chouhan; Md. Masoom Ansari; and Shivani Banjare.
The application for the patent was published on August 28, 2026, under issue no. 35/2026.
Abstract: ABSTRACT AI-Based Intelligent Phishing Detection System Using Hybrid NLP and Machine Learning Models AI-Based Intelligent Phishing Detection System Using Hybrid Natural Language Processing (NLP) and Machine Learning (ML) Models for identifying phishing attacks across diverse digital communication platforms, including emails, websites, SMS messages, instant messaging applications, and social media networks. The proposed system comprises a data acquisition module, a preprocessing engine, an NLP- based semantic analysis module, a URL intelligence module, a sender authentication module, a behavioral analytics module, a hybrid machine learning classification engine, a decision fusion module, and an automated response module. These interconnected modules collectively analyze textual, structural, behavioral, and technical attributes of digital communications to accurately distinguish phishing attempts from legitimate interactions. The system employs advanced Natural Language Processing techniques to evaluate semantic context, linguistic patterns, urgency indicators, impersonation attempts, and suspicious textual characteristics, while simultaneously extracting technical features such as domain reputation, URL structure, SSL certificate validity, email authentication records, communication metadata, and user behavioral patterns. The extracted multidimensional features are processed by a hybrid ensemble of machine learning classifiers, including Random Forest, Support Vector Machine, Gradient Boosting, Deep Neural Networks, and Transformer-based models. A decision fusion engine combines the outputs of multiple classifiers to generate a phishing confidence score, enabling highly accurate classification while reducing false positive and false negative detections. Upon detection of phishing activity, the automated response module initiates predefined cybersecurity actions including warning generation, malicious URL blocking, message quarantine, administrator notification, incident logging, and security dashboard updates. The system further incorporates a continuous learning framework that periodically retrains the NLP and machine learning models using verified phishing samples and updated cyber threat intelligence, enabling effective detection of zero-day, multilingual, and AI- generated phishing attacks.
Disclaimer: Curated by HT Syndication.
