MUMBAI, India, Jan. 9 -- Intellectual Property India has published a patent application (202521110357 A) filed by Ritu Bhadauria; Astha Patel; Bhoomi Namdev; Anuj Sharma; Archana Nair; Ayushi Awasthi; Darshani Gupta; Dipanshu Amrate; Atuj Sharma; and Atharva M Jirapure, Bhopal, Madhya Pradesh, on Nov. 12, 2025, for 'federated machine learning framework for privacy-preserving predictive analytics across distributed edge devices.'
Inventor(s) include Ritu Bhadauria; Astha Patel; Bhoomi Namdev; Anuj Sharma; Archana Nair; Ayushi Awasthi; Darshani Gupta; Dipanshu Amrate; Atuj Sharma; and Atharva M Jirapure.
The application for the patent was published on Dec. 12, under issue no. 50/2025.
According to the abstract released by the Intellectual Property India: "Federated Machine Learning Framework for Privacy-Preserving Predictive Analytics Across Distributed Edge Devices The present invention discloses a Federated Machine Learning Framework for Privacy-Preserving Predictive Analytics Across Distributed Edge Devices, designed to enable collaborative intelligence without compromising user privacy or data security. In traditional machine learning architectures, raw data must be centralized for model training, creating significant risks related to data leakage, network congestion, and regulatory non-compliance. The proposed invention eliminates these vulnerabilities by adopting a distributed learning approach, wherein individual edge devices train local models using their own data and transmit only encrypted model parameters to a central aggregator. This ensures that no raw or sensitive information leaves the local environment, maintaining data integrity and confidentiality throughout the learning process. The framework incorporates homomorphic encryption and differential privacy mechanisms to secure model updates and protect against potential inference attacks. Encrypted parameters are aggregated at the server using a secure aggregation protocol that computes the global model update without decrypting individual contributions. The aggregated model is then redistributed to all participating devices, enabling iterative improvement while ensuring that local datasets remain inaccessible to external entities. This approach significantly enhances security and privacy while maintaining model accuracy comparable to centralized systems. The invention introduces an adaptive synchronization mechanism to manage heterogeneous edge environments where devices differ in computational capacity, network bandwidth, and power constraints. By dynamically adjusting update frequencies and learning rates, the framework minimizes communication overhead and ensures balanced participation across all devices. This adaptability allows real-time analytics and predictive intelligence to be performed at the network's edge, reducing latency and improving response times in critical applications such as healthcare, smart grids, industrial automation, and autonomous systems. The result is a scalable, secure, and energy-efficient federated learning system capable of delivering advanced predictive analytics across large-scale distributed infrastructures while fully preserving user privacy."
Disclaimer: Curated by HT Syndication.
