MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611070229 A) filed by Ims Engineering College, Ghaziabad on June 04, 2026, for A System For Real-Time Deepfake Detection Using Multi-Modal Neural Networks.
Inventors include Dr. Sonali Mathur; Dr. Sonia Juneja; Dr. Nizam Uddin Khan; and Dr. Manoj Kumar Singh.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The present invention relates to a system for real-time deepfake detection using multi-modal neural networks. The system receives a digital media stream including video data, audio data, timestamp data, and metadata, and processes the stream through a pre-processing module for synchronisation and feature extraction. A visual neural analysis module detects facial artefacts, landmark instability, lighting mismatch, and frame-level generative traces, while an audio neural analysis module detects synthetic speech characteristics, voice cloning artefacts, and acoustic irregularities. A cross-modal synchronisation module compares phoneme and viseme alignment to detect lip-speech mismatch, and a temporal consistency module analyses sequential continuity across media segments. A multi-modal fusion engine combines modality- specific outputs to generate a unified deepfake probability score in real time. The system may further provide explainable suspicious-region outputs, metadata verification, biometric behaviour analysis, adaptive learning, and real-time alerts for digital identity verification, video conferencing, content moderation, cyber security, and forensic media authentication. Accompanied Drawing [FIGS. 1-2]
Disclaimer: Curated by HT Syndication.
