MUMBAI, India, July 13 -- Intellectual Property India has published a patent application (202621060188 A) filed by Prashant Sheshrao Titare; Nita Mahale; Tejas Abhay Kulkarni; Paras Vithal Yadav; Ruturaj Tatoba Mahekar; Harshal Kishor Bandgar; and Dr. Vinayak Kottawar on May 12, 2026, for Ai-Based Deepfake Detection System For Face-Swapped Video Authentication.
Inventors include Nita Mahale; Tejas Abhay Kulkarni; Paras Vithal Yadav; Ruturaj Tatoba Mahekar; Harshal Kishor Bandgar; Dr. Vinayak Kottawar; and Dr. Prashant Titare.
The application for the patent was published on July 10, 2026, under issue no. 28/2026.
Abstract: This project introduces a deepfake detection system that uses artificial intelligence (AI) and sophisticated deep learning techniques to authenticate face-swapped videos. Detecting manipulated video content has become a significant challenge in maintaining digital authenticity due to the rapid evolution of generative models. The suggested method makes use of a Bidirectional Long Short-Term Memory (BiLSTM) networkto identify temporal and spatial irregularities in video clips. After using a convolutional backbone to extract facial features from video frames, the system sequentially analyzes the data using a BiLSTM network to identify subtle artifacts and abrupt transitions. The model's ability to discern between authentic and artificially modified videos is improved by its dual-layered architecture, which allows it to analyze frame dependencies both forward and backward. The suggested BiLSTM-based method outperforms conventional CNN and unidirectional LSTM models in terms of accuracy and robustness, according to experimental evaluation on benchmark datasets like FaceForensics++ and Celeb-DF. The created system offers a dependable framework for video authentication in digital forensics and media verification applications, in addition to helping to prevent the misuse of AIgenerated media.
Disclaimer: Curated by HT Syndication.
