MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641072841 A) filed by Vellore Institute Of Technology on June 12, 2026, for Federated Learning-Based Intrusion Prediction System For Vehicular Networks.
Inventors include Santhosh Kumar Svn; and Nithin Karthikchiru Malla.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: ABSTRACT FEDERATED LEARNING-BASED INTRUSION PREDICTION SYSTEM FOR VEHICULAR NETWORKS A method (100) for intrusion prediction in a vehicular ad-hoc network comprises extracting (102) discriminative features from bidirectional network flows at a vehicle on-board unit, training (106) a local gradient boosting classifier on private local data at each vehicle node, transmitting (210) serialized model weights to a roadside unit edge aggregator, constructing (216) a weighted soft-voting ensemble by combining local gradient boosting classifiers with weights proportional to training sample counts, broadcasting (110) a global model to the vehicle nodes, maintaining (602) a rolling buffer of consecutive packets, enriching the rolling buffer with statistical summary vectors to produce an enriched feature set, feeding (614) the enriched feature set to a predictor to forecast an attack class for a next anticipated packet, and generating (116) an early intrusion alert when a predicted class probability exceeds a class-specific confidence threshold.
Disclaimer: Curated by HT Syndication.
