Intellectual Property India Publishes Patent Application for Intelligent Deep Learning Framework For Deepfake Detection And Media Integrity

MUMBAI, India, Aug. 31 -- Intellectual Property India has published a patent application (202641101608 A) filed by Chennai Institute Of Technology on August 22, 2026, for Intelligent Deep Learning Framework For Deepfake Detection And Media Integrity.

Inventors include Dr. P. Kalaiarasi; Dr T S Arulananth; Mr. R. Sarath; and Mrs. B. Gayathri.

The application for the patent was published on August 28, 2026, under issue no. 35/2026.

Abstract: The present invention pertains to an Adaptive Artificial Intelligence Framework for Real-Time Synthetic Media Authentication for the purpose of detecting and authenticating manipulated multimedia data that include images, video, and audio samples. The proposed artificial intelligence framework is built based on Convolutional Neural Networks (CNNs), transformer-based temporal analysis, and Explainable Artificial Intelligence (XAI) techniques for the identification of deepfake artifacts, anomalous motions, speech manipulations, and temporal inconsistencies in the multimedia data under consideration. The artificial intelligence framework involves multi-modal feature extraction, temporal consistency analysis, feature fusion, manipulation detection, and forensic visualization for the effective verification of the authenticity of multimedia contents. The Authenticity Intelligence Engine in the proposed artificial intelligence framework outputs the confidence scores and Media Trust Index (MTI) values to ascertain whether the multimedia data are genuine or manipulated. The adaptive artificial intelligence framework further consists of the Explainable Forensic Reporting module that is able to output heatmaps, suspicious regions localization maps, manipulation flags, and other forensic evidences to increase transparency and interpretability of the detection process. The framework is capable of performing visual and audio forensic analysis with respect to facial blending, replay attacks, voice cloning, GAN generated content, and temporal synchronization issues.

Disclaimer: Curated by HT Syndication.