MUMBAI, India, Jan. 2 -- Intellectual Property India has published a patent application (202541124183 A) filed by Sri Krishna College Of Engineering And Technology, Coimbatore, Tamil Nadu, on Dec. 9, 2025, for 'federated learning-based adaptive security system for mqtt iot devices with behavioral pattern detection and blockchain-backed model integrity ledger.'
Inventor(s) include Dr. Naveenbalaji Gowthaman; and Ms. Anusha S B.
The application for the patent was published on Jan. 2, under issue no. 01/2026.
According to the abstract released by the Intellectual Property India: "Federated Learning-Based Adaptive Security System for MQTT IoT Devices with Behavioral Pattern Detection and Blockchain-Backed Model Integrity Ledger The invention discloses a federated learning-based adaptive security framework for MQTT-based IoT devices incorporating lightweight neural anomaly detection, dynamic protocol parsing, privacy-preserving model refinement, and a novel blockchain-backed model integrity ledger. Each IoT device hosts a sub-2 MB neural network and an adaptive MQTT parser that jointly identify behavioral and structural anomalies in MQTT traffic. Upon detecting threats, devices perform localized federated learning and generate anonymized model updates. A decentralized blockchain ledger verifies the integrity of each update before cloud aggregation, preventing model poisoning attacks. Verified updates are synthesized into a global security model and redistributed across the network through real-time synchronization. This decentralized, self-learning architecture enhances privacy, ensures model authenticity, and delivers robust, scalable protection against evolving MQTT-specific threats in resource-constrained IoT environments."
Disclaimer: Curated by HT Syndication.
