MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611070286 A) filed by Sharda University on June 05, 2026, for Federated Learning Update-Screening System.
Inventors include Saumya Varshney; Mohit Kumar; Shivam Sharma; Prof. Aditya Dayal Tyagi; Mr. Mohit Kumar; and Dr. Rajneesh Kumar Singh.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: Abstract Disclosed is a federated learning update-screening system for detecting poisoned client updates before aggregation. The system comprises client computing devices that locally train respective local learning models and transmit client update data, and a federated learning update-screening gateway appliance physically interposed between the client computing devices and an aggregation server. The gateway appliance comprises an update input port, an update buffer, a screening processor, an update isolation switch, a quarantine storage unit, and a safe-update output port. The screening processor reduces dimensionality of client update data, analyses reduced update feature data for spatial irregularity, temporal deviation, and anomalous deviation, generates a combined anomaly score, and classifies each client update data as safe or malicious. The update isolation switch blocks malicious client updates and permits only safe client updates to pass toward the aggregation server for generating a global learning model update. Fig. 1 Dated June 4, 2026 / Kumar Tushar Srivastava IN/PA-3973 Agent for the Applicant
Disclaimer: Curated by HT Syndication.
