Adapting Network Information into Semantics for Generalizable and Plug-and-Play Multi-Scenario Network Diagnosis

TL;DR

This work tackles the challenge of generalizing network fault diagnosis in data-scarce, multimodal environments by introducing LNSG and NetSemantic. LNSG converts heterogeneous network states into semantic and symbolic representations, enabling LLMs to reason across modalities without bespoke fault data. A dynamic NKG with retrieval-augmented generation (RAG) supports phased planning, decomposition into subproblems, and explainable reports, yielding zero-shot fault diagnosis with strong generalization. Experimental results on a NS-3-based digital twin dataset show NetSemantic outperforming multiple baselines and maintaining robustness across topologies and network types, highlighting the approach's plug-and-play applicability and practical impact for scalable AIOps workflows.

Abstract

Leverage large language model (LLM) to refer the fault is considered to be a potential solution for intelligent network fault diagnosis. However, how to represent network information in a paradigm that can be understood by LLMs has always been a core issue that has puzzled scholars in the field of network intelligence. To address this issue, we propose LLM-based Network Semantic Generation (LNSG) algorithm, which integrates semanticization and symbolization methods to uniformly describe the entire multi-modal network information. Based on the LNSG and LLMs, we present NetSemantic, a plug-and-play, data-independent, network information semantic fault diagnosis framework. It enables rapid adaptation to various network environments and provides efficient fault diagnosis capabilities. Experimental results demonstrate that NetSemantic excels in network fault diagnosis across various complex scenarios in a zero-shot manner.

Figures (8)

• Figure 1: Overview of The Workflow of Our Proposed NetSemantic Diagnostic Framework.
• Figure 2: Mapping Rules of Network Data Time Series.
• Figure 3: Network Knowledge Graph Construction.
• Figure 4: Confusion Matrix for NetSemantic Fault Diagnosis (a) and Anomaly Detection (b). The confusion matrix shows whether the model accurately identifies true anomalous samples and has fewer errors mislabeling correct samples as anomalous.
• Figure 5: Anomaly Detection Accuracy (a) and Fault Diagnosis Accuracy (b) under networks of different scales. The comparison was performed on networks with node counts of 9-12, 13-16, and 17-20 nodes.