🤖 AI Summary
Unsupervised anomaly detection in multivariate time series under real-world conditions—where training data is contaminated with anomalous samples—remains a critical yet underexplored challenge.
Method: This paper proposes TSAD-C, an end-to-end framework that systematically models label-level noise in time-series anomaly detection for the first time. It integrates a denoising module with long-range variable dependency modeling via spatiotemporal graph neural networks and conditional diffusion models to implicitly reconstruct a clean normal manifold. An anomaly-aware denoising mechanism further enhances robustness against contaminated samples.
Contribution/Results: TSAD-C achieves state-of-the-art performance across four authoritative benchmark datasets, outperforming all existing methods. It delivers a learnable, interpretable, and highly robust solution for unsupervised time-series anomaly detection under noisy training conditions.
📝 Abstract
Mainstream unsupervised anomaly detection algorithms often excel in academic datasets, yet their real-world performance is restricted due to the controlled experimental conditions involving clean training data. Addressing the challenge of training with noise, a prevalent issue in practical anomaly detection, is frequently overlooked. In a pioneering endeavor, this study delves into the realm of label-level noise within sensory time-series anomaly detection (TSAD). This paper presents a novel and practical end-to-end unsupervised TSAD when the training data is contaminated with anomalies. The introduced approach, called TSAD-C, is devoid of access to abnormality labels during the training phase. TSAD-C encompasses three core modules: a Decontaminator to rectify anomalies (aka noise) present during training, a Long-range Variable Dependency Modeling module to capture long-term intra- and inter-variable dependencies within the decontaminated data that is considered as a surrogate of the pure normal data, and an Anomaly Scoring module to detect anomalies from all types. Our extensive experiments conducted on four reliable and diverse datasets conclusively demonstrate that TSAD-C surpasses existing methodologies, thus establishing a new state-of-the-art in the TSAD field.
