MUMBAI, India, Oct. 5 -- Intellectual Property India has published a patent application (202641114783 A) filed by Vardhaman College Of Engineering on September 25, 2026, for System And Method For Privacy-Preserving Federated Learning Across Edge Devices.
Inventors include Dr. Shaik Imam Saheb; Mr. Naveedkumarreddy K; Mr. V Shashivanth; Mr. Vankudoth Vinod Kumar; Mr. Uppala Nihar; and Ms. Bogolu Rupa.
The application for the patent was published on October 02, 2026, under issue no. 40/2026.
Abstract: ABSTRACT System and Method for Privacy-Preserving Federated Learning Across Edge Devices The present disclosure pertains to a system and method for privacy- preserving federated learning across a plurality of edge devices. The system performs local machine-learning training based on the raw training data that remains in the respective edge devices, determines sensitivity of model-parameter groups, assesses device integrity and update behaviour, and maintains device-specific privacy- budget information. A dynamic privacy-state engine generates a privacy state based on parameter sensitivity, device integrity, update risk, privacy consumption, computational capability, and communication conditions. A parameter-exposure controller selects which model parameters may be transmitted, and a privacy- transformation selector applies perturbation, masking, quantisation, sparsification, encryption, or any combination thereof. A risk-adaptive aggregation controller selects aggregation paths over which protected updates are forwarded to produce a global model. The privacy state is updated by a feedback module based on model convergence, privacy consumption, and subsequent update behaviour, thereby decreasing sensitive parameter exposure, communication overhead, computational processing, and susceptibility to anomalous updates, while enabling adaptive federated model training.
Disclaimer: Curated by HT Syndication.
