MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611070231 A) filed by Ims Engineering College, Ghaziabad on June 04, 2026, for A System For Predictive Cyber Threat Intelligence Using Graph Neural Networks.
Inventors include Dr. Sonia Juneja; Dr. Manoj Kumar Singh; Dr. Sonali Mathur; and Mr. Gaurav Gulati.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The present invention discloses a system for predictive cyber threat intelligence using graph neural networks. The system receives cyber security telemetry from multiple sources including network logs, endpoint events, authentication records, vulnerability scanners, cloud audit trails, threat intelligence feeds, and incident records. Extracted cyber entities are represented as nodes and their contextual relationships are represented as edges in a heterogeneous cyber intelligence graph. A graph neural network processing engine performs message passing and embedding generation across the graph to identify hidden threat relationships, predict malicious links, assess compromise likelihood, and forecast probable attack paths. A predictive threat scoring module assigns dynamic risk scores to entities, relationships, and subgraphs, while an explainability module identifies influential cyber evidence supporting each prediction. A response recommendation module generates mitigation actions such as blocking communication, isolating devices, disabling accounts, patching vulnerabilities, or escalating incidents. The system enables early warning, proactive cyber defence, contextual prioritisation, and reduced false positives in security operations. Accompanied Drawing [FIGS. 1-2]
Disclaimer: Curated by HT Syndication.
