Intellectual Property India Publishes Patent Application for A System And Method For Privacy-Preserving Federated Learning Across Distributed Computing Environments

MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641091427 A) filed by Pragati Engineering College on July 28, 2026, for A System And Method For Privacy-Preserving Federated Learning Across Distributed Computing Environments.

Inventors include Mrs. P. Satyavathi; Mr. Janardhana Rao Addanki; Narisetty Gopika Nischala; Dammala Hari Kishore; and Pavada Guna Sandya Deepika.

The application for the patent was published on July 31, 2026, under issue no. 31/2026.

Abstract: ABSTRACT A System and Method for Privacy-Preserving Federated Learning Across Distributed Computing Environments The present disclosure relates to a system and a method for privacy-preserving federated learning across distributed computing environments. The system includes one or more processors configured to execute a distributed federated node interface module, a hardware-anchored execution attestation module, a semantic gradient transformation module, an adaptive gradient obfuscation module, a behavioral trust learning module, a trust-weighted adaptive aggregation module, a multi-layer model integrity verification module, a poisoning attack isolation module, an adaptive federated topology optimization module, a communication compression and sparse synchronization module, a cross-layer privacy verification module, an adaptive resource orchestration module, and a secure federated model repository. The proposed system verifies trusted execution environments, transforms local model updates to privacy-preserving semantic representations, dynamically evaluates participant trustworthiness, identifies malicious model updates, optimizes collaboration topology and communication efficiency, continuously monitors cumulative privacy leakage, and intelligently allocates computation resources. Therefore, the invention improves privacy-preserving, execution integrity, attack robustness, distributed scalability, communication efficiency, and reliability of collaborative machine learning without centralized sharing of sensitive training data.

Disclaimer: Curated by HT Syndication.