MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641091354 A) filed by Sreenivasa Institute Of Technology And Management Studies on July 26, 2026, for Quantum-Enhanced Federated Deep Learning Framework For Real-Time Tamper Detection And Secure Edge-Based Image Forensics.
Inventors include Mr. A. Senthil Murugan, Assistant Professor In Department Of Cseaiml, Sreenivasa Institute Of Technology And Management Studies, Dr. D. K. Audikesavulu Marg, Bangalore-Tirupathi Bye-Pass Road; Mrs. N. Ashalatha, Assistant Professor In Department Of Cseaiml, Sreenivasa Institute Of Technology And Management Studies, Dr. D. K. Audikesavulu Marg, Bangalore-Tirupathi Bye-Pass Road, Murukambattu, Chittoor, Andhra Pradesh.; Dr. T. Senthil, Associate Professor In Department Of Cse, Sreenivasa; Mr. S Narendra Kumar, Assistant Professor In Department Of Ece; Ms. P. Poojitha, Assistant Professor In Department Of Cseaiml; and Dr. M. Dileep Kumar, Professor In Department Of Cse, Siddhartha Institute Of Technology & Sciences, Korremula Road, Narapally, Ghatkesar Mandal, Medchal Malkajgiri District, Hyderabad, Telangana..
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: The present invention relates to a Quantum-Enhanced Federated Deep Learning Framework for Real-Time Tamper Detection and Secure Edge-Based Image Forensics, developed to provide an intelligent, privacy-preserving, and scalable solution for detecting, localizing, and authenticating digital image manipulations in distributed edge computing environments. The framework integrates quantum-inspired optimization, federated deep learning, edge computing, block chain-assisted security, and explainable artificial intelligence (XAI) into a unified architecture for secure and efficient image forensic analysis. Digital images acquired from heterogeneous sources, including mobile devices, surveillance systems, cloud platforms, healthcare imaging systems, and IoT devices, are preprocessed using image enhancement, normalization, and multi-scale spatial-frequency feature extraction techniques. The extracted features are collaboratively learned through a federated deep learning engine employing Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and attention-based feature fusion without transferring raw image data, thereby preserving user privacy. Quantum-enhanced optimization algorithms improve feature selection, accelerate model convergence, optimize global parameter aggregation, and reduce communication overhead across distributed edge nodes. The framework accurately detects and localizes copy-move forgery, image splicing, object removal, resampling artifacts, AI-generated synthetic images, and deepfake manipulations in real time. A secure federated aggregation layer employing encrypted parameter exchange, differential privacy, block chain-assisted verification, and cryptographic authentication safeguards collaborative model updates against adversarial attacks and unauthorized access. Furthermore, the integrated XAI module generates interpretable attention maps, tamper localization masks, and confidence scores to support transparent forensic decision-making. Continuous adaptive learning enables the global model to evolve with emerging tampering techniques, thereby improving robustness, scalability, detection accuracy, and long-term reliability. The proposed invention provides a secure, intelligent, and high-performance platform for next-generation edge-based digital image forensics and trustworthy multimedia authentication.
Disclaimer: Curated by HT Syndication.
