MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641095326 A) filed by Vardhaman College Of Engineering on August 06, 2026, for Privacy-Preserving Federated Learning Model For Secure Data Sharing Across Institutions.
Inventors include Ms. Pallamoni S Madhavi; Ms. Sahezadi Begum; Mr. Sravan Kumar Pulla; Mr. Manney Manohar Samson; Mr. Illa Veera Venkata Durga Prasad; and Mr. T Naresh.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Privacy-Preserving Federated Learning Model for Secure Data Sharing Across Institutions is the proposed invention. The proposed invention describes a privacy- preserving federated learning model for secure data sharing across institutions by combining Split Federated Learning, Homomorphic Encryption, Trusted Execution Environments, blockchain-based governance, decentralised digital identities and artificial intelligence-driven trust evaluation. The framework we propose allows multiple institutions to jointly train machine learning algorithms without sharing raw data sets or exposing sensitive information. Local data stay within the institutional infrastructures, and encrypted intermediate feature representations and model parameters are securely aggregated to generate a global model. Homomorphic encryption lets you do operations on encrypted data , so the data is still secret during aggregation . Trusted Execution Environments are a hardware way of protecting sensitive operations . The blockchain governance layer keeps an immutable record of model updates, participant contributions, aggregation events and access permissions using intelligent contracts, providing transparency and accountability. The AI-based trust evaluation engine continuously observes the behaviour of participants, detects model poisoning attacks, evaluates the quality of contributions, and dynamically adjusts aggregation weights to improve the model trustworthiness and resilience. The privacy of the data is preserved, and the final optimised model is securely distributed to the participating institutions.
Disclaimer: Curated by HT Syndication.
