Causality-Aware Contrastive Learning for Robust Multivariate Time-Series Anomaly Detection

Utilizing the complex inter-variable causal relationships within multivariate time-series provides a promising avenue toward more robust and reliable multivariate time-series anomaly detection (MTSAD) but remains an underexplored area of research. This paper proposes Causality-Aware contrastive learning for RObust multivariate Time-Series (CAROTS), a novel MTSAD pipeline that incorporates the notion of causality into contrastive learning. CAROTS employs two data augmentors to obtain causality-preserving and -disturbing samples that serve as a wide range of normal variations and synthetic anomalies, respectively. With causality-preserving and -disturbing samples as positives and negatives, CAROTS performs contrastive learning to train an encoder whose latent space separates normal and abnormal samples based on causality. Moreover, CAROTS introduces a similarity-filtered one-class contrastive loss that encourages the contrastive learning process to gradually incorporate more semantically diverse samples with common causal relationships. Extensive experiments on five real-world and two synthetic datasets validate that the integration of causal relationships endows CAROTS with improved MTSAD capabilities. The code is available at https://github.com/kimanki/CAROTS.

Detecting anomalies in complex multivariate time series is crucial but challenging—especially when multiple factors influence each other in subtle ways. Traditional methods often overlook these interdependencies, leading to missed or incorrect detections. Our research tackles this issue by focusing on causal relationships—understanding how one variable can influence another over time. We developed a new approach called CAROTS, which teaches a machine learning model to tell apart normal and abnormal patterns by recognizing when these causal links are preserved or disrupted. To do this, we simulate both realistic and intentionally flawed versions of data, helping the model learn the difference. We also introduce a special learning method that gradually includes more diverse examples, as long as they share meaningful causal patterns. Our approach was tested on various real and synthetic datasets and showed significant improvements over existing techniques. This research opens the door to more reliable anomaly detection in critical applications where understanding cause and effect is essential.