MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202621047887 A) filed by Sahil Bagal; Suhail Hasan; Om Jangam; and Atharva Bankhele on April 15, 2026, for Neuralguard Pro: Deepfake Prevention System Architecture..
Inventors include Sahil Bagal; Suhail Hasan; Om Jangam; and Atharva Bankhele.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: A multi-modal, AI-powered deep fake detection platform is disclosed, designed to identify synthetic media through four independent forensic signals: spatial artifact detection using a fine-tuned EfficientNet-B4 convolutional neural network, frequency-domain spectral analysis via Discrete Cosine Transform GAN fingerprinting, biological liveness verification using remote photoplethysmography, and temporal consistency analysis for video inputs. A central architectural innovation termed the Multi-Signal Forensic Fusion Architecture (MSFFA) operates these four pipelines in parallel and combines their outputs through a learned weighted ensemble that adapts to input modality, producing a final tamper probability score with calibrated uncertainty. Gradient weighted Class Activation Mapping generates pixel-level spatial explainability heatmaps that visually localize suspected manipulation regions. The platform accepts JPEG, PNG, MP4, and WebM inputs, processes them through the MSFFA pipeline, and delivers a structured ForensicReport containing per pipeline evidence artifacts, visual overlays, and a human-interpretable forensic summary. The system is containerized for production deployment via Docker Compose, integrating a React frontend with an evidence-first dark-mode interface and a GPU-accelerated ML inference backend.
Disclaimer: Curated by HT Syndication.
