MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641094643 A) filed by Vardhaman College Of Engineering on August 05, 2026, for Ai-Based Detection Of Deepfake Content Using Multimodal Forensic Analysis.
Inventors include Mr. Vadla Sai Kalyan Chary; Ms. M Bhuvaneswari; Dr. Chadalavada Naga Priyanka; Mr. G Mahender; Ms. M Bharathi; and Ms. Bandari P Soundarya.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: AI-Based Detection of Deepfake Content Using Multimodal Forensic Analysis is the proposed invention. The proposed invention discloses a system for deepfake detection based on Artificial Intelligence using multimodal forensic analysis. The invention combines transformer-based deep learning, multimodal feature fusion, explainable AI, blockchain-enabled evidence management, and continual learning to reliably detect manipulated images, videos, and audio recordings. A number of digital media sources are pre-processed for frame extraction, audio synchronisation, data normalisation, and metadata analysis. Independent forensic modules analyse facial inconsistencies, speech characteristics, motion continuity, lip synchronisation, illumination patterns, compression artefacts and file integrity to extract visual, audio, temporal and metadata features. A cross-modal transformer network combines these heterogeneous features through adaptive attention mechanisms to generate a reliable authenticity assessment with high confidence. To produce interpretable forensic reports, an explainable AI module is proposed, which includes anomaly localisation and confidence estimation, and modality-specific visualisations that allow transparent decision-making for investigators and authorities responsible for regulation. Tamper-proof recording of verified forensic results and cryptographic hashes in a permissioned blockchain guarantees the evidence preservation, traceability, and legal compliance. A continual learning framework allows the model to adjust to the emerging deepfake generation techniques.
Disclaimer: Curated by HT Syndication.
