CGSTA: Cross-Scale Graph Contrast with Stability-Aware Alignment for Multivariate Time-Series Anomaly Detection

2026-02-24

AI Summary

This work addresses the challenges of multivariate time series anomaly detection, which is often compromised by evolving dynamic dependencies and noise, leading to false positives or missed detections. To this end, we propose a robust detection method based on multi-scale dynamic graphs: it models time-varying inter-variable dependencies through a dynamically hierarchical graph structure, enhances structural awareness via cross-scale graph contrastive learning, and incorporates a stability-aware alignment mechanism to suppress noise. Furthermore, multi-scale temporal features are effectively fused, and precise anomaly scores are generated through conditional density estimation. Extensive experiments on four benchmark datasets demonstrate that the proposed method achieves state-of-the-art performance on PSM and WADI, matches the best existing approaches on SWaT and SMAP, and significantly improves both detection accuracy and robustness.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Time-Series/Data Streams
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Multivariate time-series anomaly detection is essential for reliable industrial control, telemetry, and service monitoring. However, the evolving inter-variable dependencies and inevitable noise render it challenging. Existing methods often use single-scale graphs or instance-level contrast. Moreover, learned dynamic graphs can overfit noise without a stable anchor, causing false alarms or misses. To address these challenges, we propose the CGSTA framework with two key innovations. First, Dynamic Layered Graph Construction (DLGC) forms local, regional, and global views of variable relations for each sliding window; rather than contrasting whole windows, Contrastive Discrimination across Scales (CDS) contrasts graph representations within each view and aligns the same window across views to make learning structure-aware. Second, Stability-Aware Alignment (SAA) maintains a per-scale stable reference learned from normal data and guides the current window's fast-changing graphs toward it to suppress noise. We fuse the multi-scale and temporal features and use a conditional density estimator to produce per-time-step anomaly scores. Across four benchmarks, CGSTA delivers optimal performance on PSM and WADI, and is comparable to the baseline methods on SWaT and SMAP.

Problem

• multivariate time-series
• anomaly detection
• dynamic graph
• noise robustness
• inter-variable dependencies

Innovation

• Dynamic Layered Graph Construction
• Contrastive Discrimination across Scales
• Stability-Aware Alignment
• Multivariate Time-Series Anomaly Detection
• Graph Contrastive Learning