MUMBAI, India, Oct. 12 -- Intellectual Property India has published a patent application (202641118754 A) filed by Easwari Engineering College; and Srm Institute Of Science And Technology, Ramapuram Campus on October 05, 2026, for Ai-Based Attention-Driven Graph Neural Network System For Detecting And Explaining Intrusions In Iot Networks.
Inventors include Barani M S; Tamilarasi L; and Preethi B.
The application for the patent was published on October 09, 2026, under issue no. 41/2026.
Abstract: An intelligent intrusion detection and explanation model using graphs in artificial intelligence is provided in order to detect and explain intrusions within IoT-based networks. The traffic flows within the network are modelled as nodes in the graph, where they are related based on their behavioural similarity. An attention-based neural encoder is used to analyse the graph by attributing varying weights to the links as opposed to considering all neighbouring flows equally, hence making the system more effective at detecting infrequent attack patterns. A training model that dynamically adapts the balance of the often occurring and less occurring intrusion types is employed to address the degradation problem caused by unbalanced network traffic data. For each detected intrusion, the invention calculates the importance level of various features within the traffic flow through gradient analysis, and provides the reasoning behind the decision as the feature importance scores. Total Number of Words in Abstract: 146
Disclaimer: Curated by HT Syndication.
