MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202621096423 A) filed by Symbiosis Skills And Professional University on August 10, 2026, for Closed Loop Node Reputation And Adaptive Privacy Budget Allocation In Federated Learning Networks.
Inventors include Rudresh Shirwaikar; Jai Mali; and Parth Kapare.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The invention discloses a computer-implemented system and method for closed-loop federated learning wherein node reputation scores are utilised for adaptive privacy-budget allocation and aggregation weight determination. The system (100) comprises a plurality of nodes (110) and at least one aggregation server (120). Each node (110) is configured with a sensitivity classifier (111), a local model trainer (112), a reputation-scoring module (113), a differential-privacy mechanism (114), and a ledger client (115). The aggregation server (120) comprises a tier-consistency verification module (121), a closed-loop control module (122), a global model updater (123), and a ledger coordinator (124). The method (200) comprises steps of classifying datasets into sensitivity tiers (210), computing composite reputation scores (220), allocating epsilon values adaptively (230), verifying tier consistency (240), overriding privacy parameters (250), and recording decisions in a cryptographically verifiable audit ledger (400). The invention ensures enforceable privacy guarantees, regulatorily auditable aggregation outcomes, reduction of unnecessary noise injection, and demonstrable industrial applicability across regulated sectors.
Disclaimer: Curated by HT Syndication.
