MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641106493 A) filed by St. Peters Engineering College on September 03, 2026, for A Federated Transformer Learning Framework With Differential Privacy For Distributed Intelligent Analytics.
Inventors include Mr. A. D. Sivarama Kumar, Assistant Professor In Department Of Cse, St; Mr. Rathod Koveed, Assistant Professor In Department Of Cse, St Peters Engineering College, Opposite Ts Forest Academy, Kompally Road, Dullapally, Maisammaguda, Medchal, Hyderabad, Telangana; Mr. David Boon Merugumalla, Assistant Professor In Department Of Cse; Mr. K. Avinash, Assistant Professor In Department Of Cseaiml, St; Mrs. M Karuna Sudha, Assistant Professor In Department Of Mba, St; and Mrs. Muppalaneni Subhashini, Assistant Professor In Department Of Cseaiml, St Peters Engineering College, Opposite Ts Forest Academy Kompally Road, Dullapally, Maisammaguda, Medchal, Hyderabad, Telangana.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The invention discloses a Federated Transformer Learning Framework that integrates transformer architectures, federated optimization techniques, and differential privacy mechanisms to enable secure, distributed intelligent analytics across decentralized data sources. Traditional centralized machine learning approaches require raw data aggregation, which raises significant privacy and compliance concerns in sensitive domains such as healthcare, finance, and IoT. The proposed system eliminates the need for data pooling by allowing multiple client nodes to collaboratively train transformer based models while retaining data locally. The framework consists of three core components: local client nodes that perform transformer training on private datasets, a federated aggregation server that securely combines model updates, and a differential privacy module that injects calibrated noise into gradients or parameters before transmission. This design ensures that individual contributions remain indistinguishable, thereby safeguarding sensitive information against adversarial inference. By leveraging the representational power of transformers, the invention achieves superior performance in tasks involving text, images, and multimodal data. Communication efficient aggregation strategies reduce bandwidth consumption, while adaptive optimization techniques enhance robustness in non IID data environments. The system is scalable, resilient, and compliant with global data protection standards. This invention provides a transformative solution for industries requiring both advanced analytics and strict privacy preservation. It enables secure collaboration across distributed data silos, delivering high accuracy insights in domains such as cross hospital medical research, financial fraud detection, and large scale IoT analytics.
Disclaimer: Curated by HT Syndication.
