MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202641061024 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on May 14, for 'deepfake detection system using multi-domain forensic features with convolutional and transformer-based architectures.'
Inventor(s) include Muthunagai; Barath Selvakumar; Gokul Raj P; and Pranesh G.
The application for the patent was published on May 29, under issue no. 22/2026.
According to the abstract released by the Intellectual Property India: "A deepfake detection system (106) for detecting manipulated digital images is disclosed. The system comprises a processor (108) and a memory (110) storing instructions executable by the processor. The processor is configured to receive an input image and extract multi-domain forensic features from the image. The extracted features are combined into a unified multi-channel representation configured to capture spatial, structural, frequency-domain, and compression-based characteristics. The unified representation is processed using a convolution-based architecture to obtain local feature representations and a transformer-based architecture to obtain global contextual representations. The local and global representations are fused using a cross-attention mechanism and refined using an artifact-aware attention mechanism. The refined representations are used to classify the input image as real or fake. The system provides improved accuracy and robustness in detecting manipulated images by leveraging multi-domain feature fusion and hybrid feature learning."
Disclaimer: Curated by HT Syndication.
