Adversarial-Robust Multivariate Time-Series Anomaly Detection via Joint Information Retention

2026-03-26

AI Summary

This work addresses the limited robustness of multivariate time series anomaly detection models, which often suffer from sensitivity to local perturbations and structured noise. To mitigate this issue, the authors propose ARTA, a framework that jointly trains an anomaly detector with a sparsity-constrained mask generator through a min-max adversarial optimization objective, thereby enhancing model stability under perturbations. A key innovation is the introduction of an interpretable adversarial masking mechanism that explicitly identifies temporally sensitive regions and steers the detector toward more robust global patterns rather than spurious local artifacts. Experimental results demonstrate that ARTA significantly outperforms existing methods on the TSB-AD benchmark and multiple datasets, while exhibiting notably smoother performance degradation under noise-augmented conditions.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Time-series anomaly detection (TSAD) is a critical component in monitoring complex systems, yet modern deep learning-based detectors are often highly sensitive to localized input corruptions and structured noise. We propose ARTA (Adversarially Robust multivariate Time-series Anomaly detection via joint information retention), a joint training framework that improves detector robustness through a principled min-max optimization objective. ARTA comprises an anomaly detector and a sparsity-constrained mask generator that are trained simultaneously. The generator identifies minimal, task-relevant temporal perturbations that maximally increase the detector's anomaly score, while the detector is optimized to remain stable under these structured perturbations. The resulting masks characterize the detector's sensitivity to adversarial temporal corruptions and can serve as explanatory signals for the detector's decisions. This adversarial training strategy exposes brittle decision pathways and encourages the detector to rely on distributed and stable temporal patterns rather than spurious localized artifacts. We conduct extensive experiments on the TSB-AD benchmark, demonstrating that ARTA consistently improves anomaly detection performance across diverse datasets and exhibits significantly more graceful degradation under increasing noise levels compared to state-of-the-art baselines.

Problem

• adversarial robustness
• multivariate time-series
• anomaly detection
• structured noise
• input corruption

Innovation

• adversarial robustness
• multivariate time-series anomaly detection
• joint training
• min-max optimization
• explainable AI