MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202641092128 A) filed by Srm University- Ap on July 29, 2026, for System And Method For Privacy-Preserving Federated Learning In Heterogeneous Iot Environments.
Inventors include Mellamarthi Mounika; Ammar Summaq; and Dr. Sunil Chinnadurai.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: A privacy-preserving federated learning system (100) for heterogeneous Internet of Things environments. This three-tier, privacy-preserving federated learning system (100) is designed for heterogeneous IoT environments. At its core, a central server (102) manages the global model, executes global aggregation, and utilizes a staleness weight adjustment module (112) to manage delayed updates. Connecting the system together, multiple edge servers (104) perform intermediate model aggregation for designated fleets (108) of client devices (106). To secure decentralized data, each client device is equipped with a privacy layer (110) featuring a gradient transformation module (114). This module defends against reconstruction attacks by strategically altering the direction of gradients while strictly preserving their magnitude, successfully maintaining model convergence while guaranteeing robust data privacy. (FIG. 1)
Disclaimer: Curated by HT Syndication.
