This research establishes a comprehensive taxonomy for improving Graph Neural Networks (GNNs) using Large Language Models (LLMs) to enhance trustworthiness.

• Integrates LLMs with GNNs to improve semantic understanding and generation capabilities
• Provides a structured framework for researchers to comprehend principles and applications
• Addresses critical security concerns in graph-based AI systems
• Offers practical approaches for building more reliable and secure GNN models

For security professionals, this research matters because it systematically addresses trustworthiness challenges in graph-based AI systems that are increasingly deployed in sensitive applications like fraud detection, network security, and social network analysis.

Trustworthy GNNs with LLMs: A Systematic Review and Taxonomy