MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641049267 A) filed by Sr University, Warangal, Telangana, on April 17, for 'an intelligent system for real-time phishing detection using neural network models.'
Inventor(s) include Dr. Arivukarasi M; and Dr Balajee Maram.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "The present invention relates to the development of an intelligent system for real-time detection of phishing attacks using state-of-the-art neural network models. The system analyzes received emails, URLs, and web pages using a hierarchical model based on CNNs for extracting features from textual and structural data, RNNs with LSTM components for analyzing the user's behavior sequentially, and an attention mechanism based on transformers for embedding contexts. Real-time inference is performed using lightweight models trained using quantization and knowledge distillation techniques and deployed at the edge to provide sub-millisecond performance even on devices with limited resources. The system uses a dynamically changing threat database that is continuously updated using federated learning. Benchmark tests using popular databases (such as the Phishing URL Dataset and the Enron Corpus) show the accuracy of 98.5%, precision of 99.2%, and a false positive rate of 0.2%, outperforming conventional systems by 15-20%."
Disclaimer: Curated by HT Syndication.
