MUMBAI, India, Sept. 7 -- Intellectual Property India has published a patent application (202611084186 A) filed by Upes on July 09, 2026, for A System And Method For Privacy-Preserving Federated Learning Using An Asic Accelerator For Iot Edge Devices..
Inventors include Sushabhan Choudhury; Keshav Sinha; and Ayush Kumar Agrawal.
The application for the patent was published on September 04, 2026, under issue no. 36/2026.
Abstract: The invention relates to edge artificial intelligence hardware and particularly to an ASIC-based system and method for privacy-preserving federated learning at an IoT edge node. A sensor interface receives local sensor data or feature streams and stores them in on-chip memory including an input feature buffer. A local-training accelerator having a fixed-point or low-precision multiply-accumulate array, an optimizer controller, and a micro-batch controller performs compact on-chip local update intervals using selected model parameters or low-rank adaptation parameters. Generated gradients or model deltas are processed by a gradient compression module that retains salient information and encodes a compressed update packet. A secure update engine protects the packet, and an energy-aware communication scheduling module controls selective wireless transmission to a coordinator, so that raw sensor data remains local to the device.
Disclaimer: Curated by HT Syndication.
