MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641065105 A) filed by Srm Trp Engineering College on May 23, 2026, for Context-Aware Voice Phishing Detection Through Speech-To-Text And Bert Modeling.
Inventors include Mrs. A Amalorpavam; Mohan P; Monika A; Narmadha R; and Prathap P.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: The present invention introduces a context-aware voice phishing detection system designed to help users identify fraudulent voice calls in real time. This innovative system captures live audio from phone calls and converts spoken words into text using advanced speech recognition technology. Once the speech is converted to text, the system analyzes it using a lightweight BERT-based Natural Language Processing (NLP) model. Unlike traditional systems that rely on simple keyword matching, this approach understands the context of the conversation, allowing it to detect phishing intent more accurately. To reduce false positives, the system includes a contact trust analysis mechanism that assigns trust scores based on the caller's identity. This means that if the caller is someone you know, the system is less likely to raise an alarm. Based on the phishing probability and the trust score, a dynamic risk evaluation mechanism determines the threat level. This ensures that users receive alerts only when necessary, minimizing unnecessary disruptions. When a potential phishing threat is detected, the system provides instant notifications through intelligent haptic feedback. Users receive distinct vibration patterns on their phones, letting them know the nature and severity of the threat without having to look at their screens.
Disclaimer: Curated by HT Syndication.
