MUMBAI, India, March 13 -- Intellectual Property India has published a patent application (202521132783 A) filed by Vijeet Meshram; and Pooja Kavishwar, Nagpur, Maharashtra, on Dec. 29, 2025, for 'system and method for synthetic media detection via gradient-response decay profiling and adversarial perturbation stability analysis.'
Inventor(s) include Vijeet Meshram; and Pooja Kavishwar.
The application for the patent was published on March 13, under issue no. 11/2026.
According to the abstract released by the Intellectual Property India: "The present invention discloses a universal method for distinguishing authentic sensor-captured images from algorithmically generated synthetic media. Unlike conventional detectors that search for visual artifacts, the proposed system analyzes the "Adversarial Robustness" of the image. The method subjects an image to a micro-perturbation using gradient-based noise (Epsilon=0.01) and measures the "Decay Rate" of its latent feature vectors extracted via a VGG16 CNN. The invention exploits the finding that synthetic images, derived from low-dimensional latent spaces, exhibit significantly higher structural fragility (Avg Decay: 29.49) compared to the high-entropy stochastic noise patterns inherent in physical camera sensors (Avg Decay: 28.15). The system provides a lightweight, compression-robust forensic tool."
Disclaimer: Curated by HT Syndication.
