MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611070230 A) filed by Ims Engineering College, Ghaziabad on June 04, 2026, for A Method For Adaptive Federated Learning With Privacy-Preserving Data Aggregation.
Inventors include Dr. Amit Chugh; Dr. Sonia Juneja; Dr. Nizam Uddin Khan; and Dr. Vaishali Bhargava.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The present invention relates to a method for adaptive federated learning with privacy-preserving data aggregation. The method includes initializing a global machine learning model at a central orchestration server, adaptively selecting client nodes based on data representativeness, reliability, communication capability, privacy budget and contribution history, and enabling each selected client node to locally train the global model using locally retained data. Each client node generates a local model update and transforms the update using privacy-preserving mechanisms including clipping, noise addition, secure masking, encryption, secret sharing, compression or quantization. The server evaluates received protected updates for anomaly and contribution quality, computes adaptive aggregation weights based on reliability, privacy noise, update freshness and learning contribution, and updates the global model using weighted secure aggregation without accessing raw client data. The method further supports dropout-tolerant aggregation, communication-efficient synchronization and audit logging for privacy-governed collaborative artificial intelligence training. Accompanied Drawing [FIGS. 1-2]
Disclaimer: Curated by HT Syndication.
