MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641048378 A) filed by Dr. S Angel Latha Mary; Catherine Esther Karunya A; Pavarithikesh R; Rakshitha B; Ramana Sree S; Rithika S; Rosun Ram B; Sabarish Vishal M; Samyuktha R V; and Nighash T, Coimbatore, Tamil Nadu, on April 16, for 'multimodal deep fake detection system.'
Inventor(s) include Dr. S Angel Latha Mary; Catherine Esther Karunya A; Pavarithikesh R; Rakshitha B; Ramana Sree S; Rithika S; Rosun Ram B; Sabarish Vishal M; Samyuktha R V; and Nighash T.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "The present invention relates to a multimodal deep fake detection system designed to identify manipulated or synthetic multimedia content using an integrated and intelligent analytical framework. The system processes multiple data modalities, including video, image, and audio, to enhance detection accuracy and overcome the limitations of traditional single-modality approaches. The proposed system comprises an input module that accepts multimedia data, followed by a preprocessing unit that performs operations such as frame extraction, audio extraction, noise removal, and normalization. The system then utilizes parallel feature extraction modules, where visual features are analyzed using convolutional neural networks (CNN), and audio features are extracted using techniques such as Mel-frequency cepstral coefficients (MFCC) and recurrent neural networks (RNN). These modules identify anomalies in facial structures, expressions, voice patterns, and temporal synchronization. A multimodal fusion layer integrates the extracted features to capture cross-modal inconsistencies, such as mismatches between lip movements and speech. The fused data is then processed by a classification module, which determines whether the input content is authentic or manipulated and generates a confidence score. The system further supports scalability and real-time processing through optimized algorithms and cloud-based deployment. It may also include explainability mechanisms that provide interpretable insights into detection results. Security features such as data protection and robustness against adversarial inputs are incorporated to ensure reliability. Overall, the invention provides a comprehensive, accurate, and scalable solution for deep fake detection, enabling improved digital content verification, enhanced cybersecurity, and mitigation of misinformation in modern multimedia environments."
Disclaimer: Curated by HT Syndication.
