CARE: A Cascaded Framework for Efficient and Reliable Time Series Anomaly Detection

2026-08-03

AI Summary

This work addresses the high inference cost and inefficiency of existing deep time series anomaly detection methods, which uniformly process all samples regardless of their anomaly likelihood. To overcome this limitation, the authors propose CARE, a model-agnostic framework that cascades a lightweight pre-filtering model (LPM) with a high-capacity detection model (CDM). The LPM employs a residual MLP autoencoder and a normality-conditioned gating mechanism to rapidly filter out high-confidence normal samples, forwarding only ambiguous instances to the CDM for refined discrimination. Additionally, CARE incorporates a structural attention module to explicitly model channel-wise anomaly contributions and introduces a confidence-guided selective routing objective to optimize the gating network. Evaluated on eight real-world benchmarks, CARE achieves 2.7–4.8× faster inference while maintaining detection performance on par with state-of-the-art methods.

Abstract

While deep learning models have achieved state-of-the-art performance in time series anomaly detection, their complex architectures incur substantial inference overhead. Existing methods typically apply a uniform inference strategy across all data points, which is inefficient given that anomalies are inherently scarce and the vast majority of temporal data consists of predictable normal patterns. To mitigate this bottleneck, we propose CARE, a model-agnostic cascaded inference framework that integrates a Lightweight Pre-filter Model (LPM) with an existing high-capacity Complex Detection Model (CDM). The LPM rapidly filters high-confidence normal samples using a Residual MLP AutoEncoder and a Normality-Conditioned Gating mechanism. Crucially, we introduce a Structure Attention module to explicitly capture channel-wise anomaly contributions, and optimize the gating network via a confidence-guided selective routing objective that learns reliable routing decisions to reduce unnecessary CDM invocations. Extensive experiments across eight real-world benchmarks demonstrate that CARE effectively isolates high-confidence normal samples. By routing only uncertain samples to the CDM, our framework achieves $2.7\times$ to $4.8\times$ inference speedup compared to the most accurate SOTA approaches, while still maintaining competitive detection quality.

Problem

• time series anomaly detection
• inference efficiency
• model complexity
• normal pattern filtering
• computational overhead

Innovation

• cascaded inference
• lightweight pre-filter
• structure attention
• confidence-guided routing
• time series anomaly detection