MUMBAI, India, Sept. 7 -- Intellectual Property India has published a patent application (202641104402 A) filed by Srm Trp Engineering College on August 31, 2026, for A Reliability-Aware Multimodal Framework For Robust Deepfake Detection Using Weighted Late Fusion.
Inventors include Harsika S; Priyaa Ravi; and Perumalraja Rengaraju.
The application for the patent was published on September 04, 2026, under issue no. 36/2026.
Abstract: The present invention relates to a multimodal deepfake detection system and method for identifying manipulated or synthetically generated multimedia content through integrating forensic analysis of visual, audio, physiological, spectral, and metadata-based features.The system uses a Vision Transformer (ViT-B/16) architecture for visual forgery analysis, a Wav2Vec2-based architecture for speech context analysis, an rPPG-based block for physiological consistency evaluation, an FFT-based block for detection of spectral artifacts, and a metadata analysis block for identifying encoding and compression inconsistencies. Multimodal feature representations extracted are combined through an uncertainty-aware adaptive fusion framework to produce a final deepfake classification result. The invention presented also brings improvements in the accuracy of the detection, robustness, reliability, and the capability of a cross-domain generalization of the method against advanced deepfake generation and multimedia manipulation techniques encountered in real-world forensic environments.
Disclaimer: Curated by HT Syndication.
