MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641113743 A) filed by Koneru Lakshmaiah Education Foundation Deemed To Be University on September 23, 2026, for System And Method For Malicious Package Detection In Software Dependency Networks Using Graph Neural Networks.
Inventors include Dr. Jagan Mohan Reddy Danda; and Payala Krishnanjaneyulu.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: A system (100) and method for detecting malicious software packages in dependency networks using Graph Neural Networks is disclosed. The system comprises a data collection module (101) for retrieving package data from npm, PyPI and GitHub; a graph construction module (102) for building heterogeneous dependency graphs; a feature engineering module (103) for extracting node features; a Graph Attention Network (GAT) module (104) with multi-headed attention for learning structural representations; a temporal pattern integration module (105) for detecting anomalous update patterns; a classification module (106) for outputting risk scores; and an explainability module (107) for generating interpretable explanations. The Dep-GAT architecture incorporates multiple graph attention layers, temporal integration, and dense layers with dropout regularization. The system achieves over 92% accuracy and a false positive rate below 5%, outperforming baseline methods by at least 15%. The invention addresses the critical need for detecting zero-day malicious packages in the software supply chain and provides explainable results for security teams.
Disclaimer: Curated by HT Syndication.
