Large Language Model(LLM) assisted End-to-End Network Health Management based on Multi-Scale Semanticization

TL;DR

The paper tackles network health management in heterogeneous networks by addressing multi-scale data and lifecycle gaps left by traditional methods. It introduces MSADM, a Multi-Scale Semanticized Anomaly Detection Model that unifies heterogeneous KPI scales via a dynamic rule base and semantic rule trees, paired with an attention-based detector. An LLM-based end-to-end framework then analyzes detection results, generates detailed fault analyses, and outputs actionable mitigation guidance, including executable scripts. Experimental results on NS-3 simulations and the UNSW-15 dataset show MSADM achieving high anomaly-detection accuracy ($$97.09\%$$) and fault-diagnosis accuracy ($$89.42\%$$), with semanticization markedly improving the quality and usefulness of generated maintenance reports, demonstrating practical impact for autonomous and assisted network operations.

Abstract

Network device and system health management is the foundation of modern network operations and maintenance. Traditional health management methods, relying on expert identification or simple rule-based algorithms, struggle to cope with the dynamic heterogeneous networks (DHNs) environment. Moreover, current state-of-the-art distributed anomaly detection methods, which utilize specific machine learning techniques, lack multi-scale adaptivity for heterogeneous device information, resulting in unsatisfactory diagnostic accuracy for DHNs. In this paper, we develop an LLM-assisted end-to-end intelligent network health management framework. The framework first proposes a Multi-Scale Semanticized Anomaly Detection Model (MSADM), incorporating semantic rule trees with an attention mechanism to address the multi-scale anomaly detection problem in DHNs. Secondly, a chain-of-thought-based large language model is embedded in downstream to adaptively analyze the fault detection results and produce an analysis report with detailed fault information and optimization strategies. Experimental results show that the accuracy of our proposed MSADM for heterogeneous network entity anomaly detection is as high as 91.31\%.

Figures (10)

• Figure 1: Packet Loss Rate Distribution of Network Nodes.
• Figure 2: Heterogeneous Network Health Management Scheme Architecture.
• Figure 3: Multi-scale Anomaly Detection Model Architecture.
• Figure 4: Anomaly Report Generation Process.
• Figure 5: Prompt Structure.