CAAD: Causality-Aware Multivariate Time Series Anomaly Detection via Multi-Scale Alignment and Structural Causal Consistency

Jul 16, 2026

Operational integrity of complex industrial systems depends on accurate multivariate time-series anomaly detection and diagnosis. Prior approaches often emphasize temporal similarity in learned representations, overlooking disruptions in internal causal relationships that signal failures and latent anomalies. This paper introduces CAAD, reframing anomaly detection as continuous verification of Granger-causality consistency using exogenous variables. CAAD models exogenous signals as residuals, flags anomalies via deviations caused by external interventions, and uses multi-scale alignment plus a gradient-based matrix to monitor causal breakdowns. Experiments on real industrial datasets show strong high-precision performance over state-of-the-art baselines.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

Introduction

Motivation and limitations of magnitude-based and correlation-centric detectors

Relation to causality-aware anomaly detection and Granger causality

CAAD: Causality-Aware Multivariate Time Series Anomaly Detection via Multi-Scale Alignment and Structural Causal Consistency Xin Wang [xin.wang.9@stonybrook.edu](xin.wang.9@stonybrook.edu)[ ](xin.wang.9@stonybrook.edu)Stony Brook University Stony Brook, New York, USA Yunshi Wen [yunshi-wen@outlook.com](yunshi-wen@outlook.com)[ ](yunshi-wen@outlook.com)Rensselaer Polytechnic Institute Troy, New York, USA Yanan He [yanan.he@yale.edu](yanan.he@yale.edu) Yale University New Haven, Connecticut, USA Haotian Xu [haotian.xu@stonybrook.edu](haotian.xu@stonybrook.edu)[ ](haotian.xu@stonybrook.edu)Stony Brook University Stony Brook, New York, USA Youlan Zhao [youlan.zhao@stonybrook.edu](youlan.zhao@stonybrook.edu)[ ](youlan.zhao@stonybrook.edu)Stony Brook University Stony Brook, New York, USA Michel Ferreira Cardia Haddad [m.haddad@qmul.ac.uk](m.haddad@qmul.ac.uk)[ ](m.haddad@qmul.ac.uk)Queen Mary University of London London, United Kingdom arXiv :2607 .08555v 1 [ cs .LG] 9 Jul 2026 Tengfei Ma [tengfei.ma@stonybrook.edu](tengfei.ma@stonybrook.edu)[ ](tengfei.ma@stonybrook.edu)Stony Brook University Stony Brook, New York, USA Abstract The operational integrity of complex industrial systems relies on precise anomaly detection and diagnosis. The vast majority of existing methods narrowly focus on capturing temporal similarities of representations, often overlooking the disruption of internal causal relationships, which characterizes system failures and latent anomalies. In this paper, we propose a novel framework (CAAD) that reframes anomaly detection as the continuous verification of Granger causality consistency through exogenous variables. Specifically, the CAAD framework models exogenous time-series variables as residuals, identifying anomalies as significant deviations caused by external interventions. The proposed framework leverages multi-scale alignment to internalize system dynamics and utilizes a gradient-based matrix to monitor internal causal relationship breakdowns. By quantifying causal deviations of both dynamic evolution and relational topology, the CAAD is able to capture subtle causal shifts to achieve precise anomaly detection. Extensive experiments on real-world industrial datasets demonstrate that the CAAD achieves high-precision anomaly detection, outperforming most state-of-the-art baselines. CCS Concepts • Computing methodologies → Machine learning. Keywords Time Series; Anomaly Detection; Causality; Multi-Scale Alignment This work is licensed under a Creative Commons Attribution-NonCommercialNoDerivatives 4 .0 International License. KDD’26, Jeju Island, Republic of Korea © 2026 Copyright held by the owner/author(s) . ACM ISBN 979-8-4007-2259-2/2026/08 [https://doi.org/10.1145/3770855.3817978](https://doi.org/10.1145/3770855.3817978) ACM Reference Format: Xin Wang, Yunshi Wen, Yanan He, Haotian Xu, Youlan Zhao, Michel Ferreira Cardia Haddad, and Tengfei Ma. 2026. CAAD: Causality-Aware Multivariate Time Series Anomaly Detection via Multi-Scale Alignment and Structural Causal Consistency. In Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD’26), August 09–13, 2026, Jeju Island, Republic of Korea. ACM, New York, NY, USA, 11 pages. [https:](https:)//[doi.org/10.1145/3770855.3817978](doi.org/10.1145/3770855.3817978) 1 Introduction The reliable operation of complex industrial systems, such as smart grids, water treatment facilities, and large-scale cloud infrastructures, critically depends on accurate multivariate time series anomaly detection. Failures in these systems can propagate rapidly and lead to severe safety, economic, or societal consequences, making early and reliable anomaly identification indispensable [39, 51] . In recent years, deep learning methods have been increasingly explored for this task, with approaches largely centered around reconstruction, prediction, or representation association objectives [

FAQs

What problem does CAAD address in multivariate time-series anomaly detection?

CAAD targets anomalies caused by breakdowns in internal causal logic, which may occur before large-amplitude numerical changes.

How does CAAD detect anomalies in terms of causal relationships?

CAAD reframes detection as continuous verification of Granger causality consistency using exogenous variables modeled as residuals, treating significant deviations induced by external interventions as anomalies.

What techniques does CAAD use to monitor causal consistency over time and resolutions?

CAAD uses multi-scale alignment to internalize system dynamics and a gradient-based matrix to track causal relationship breakdowns in both dynamic evolution and relational topology.

Jul 16, 2026

Upload by Valentina