MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641072851 A) filed by Vellore Institute Of Technology on June 12, 2026, for Federated Deep Learning Framework For Network Intrusion Detection.
Inventors include Santhosh Kumar Svn; and Sheetal Reddy.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: ABSTRACT FEDERATED DEEP LEARNING FRAMEWORK FOR NETWORK INTRUSION DETECTION A method (200, 300, 400, 600, 700) for federated network intrusion detection comprises receiving network traffic records at a data preprocessing phase (104), applying label encoding to categorical features, standard scaling normalisation, and synthetic minority oversampling to produce balanced training samples. The method comprises applying a multi-objective feature optimisation phase (106) comprising a non-dominated sorting genetic algorithm with binary chromosome encoding to minimise feature count and maximise detection accuracy through Pareto-optimal selection, producing a feature-selected dataset. The method comprises distributing the feature-selected dataset across client devices in a federated training phase (108), wherein each client device trains a neural network model (500) on a local partition without transmitting raw data across client boundaries, aggregating model weights at a federated aggregation phase (110) using federated averaging to produce a global model, and evaluating the global model at an evaluation phase (112) using explainability and zero-day simulation.
Disclaimer: Curated by HT Syndication.
