MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109861 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 13, 2026, for A Deep Learning-Based Framework For Automated Deepfake Video Detection And Digital Media Authentication.
Inventor includes Mr. Bhupesh Deka.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: ABSTRACT [0028] Deepfake videos that replace one individual’s face with another have introduced both creative possibilities and serious risks to privacy, media integrity, and cybersecurity. Existing detection approaches often lag behind the rapid evolution of sophisticated manipulation techniques. This project develops a hybrid deep learning framework that utilizes the Deepfake Detection dataset together with ResNeXt, Long Short-Term Memory networks, YOLOv8, and a discriminator module. The system extracts full-frame spatial features, models temporal sequences across frames, and isolates facial regions to uncover subtle artifacts introduced during fabrication. By fusing these complementary analyses, the framework achieves high accuracy in classifying videos as real or fake. Performance is validated primarily through accuracy metrics, confirming the reliability of the approach in addressing contemporary deepfake challenges.
Disclaimer: Curated by HT Syndication.
