Future of Time Series Anomaly Detection: From Historical Review to Unified Outlook
Authors: Zhijie Zhong, Zhiwen Yu, Pengyu Li, Liujian Zhang, Qiaoli Ni, Jiang Jun, Jianming Lv, Kaixiang Yang, C. L. Philip Chen
Organizations: School of Future Technology, South China University of Technology, Guangzhou, 510650, China · Pengcheng Laboratory, Shenzhen, 518066, China · School of Computer Science and Engineering, South China University of Technology, Guangzhou, Guangdong 510650, China · Pengcheng Laboratory, Shenzhen, Guangdong 518066, China · HUAWEI Technologies Co., Ltd., Shenzhen, Guangdong 518066, China · Institute of Computing Technology Chinese Academy of Sciences, South China University of Technology, Guangzhou, Guangdong 510641, China
Time series anomaly detection (TSAD) is a foundational problem in data mining, aiming to identify observations that deviate significantly from nominal temporal patterns. Driven by escalating data volumes and application complexity, recent research has yielded diverse architectures, including diffusion models, state space models, and large language models (LLMs). However, existing surveys provide limited coverage of these emerging designs and lack a unified treatment of multivariate anomalies, cross-domain unified modeling, metric reliability, and contemporary evaluation challenges. To bridge these gaps, this survey delivers three core contributions: (1) We establish a refined anomaly categorization and a structured taxonomy of paradigms reflecting modern advances, comparing detection mechanisms and charting developmental trajectories across paradigms. (2) We provide an extensive analysis of evaluation methodologies and conduct comprehensive experiments to assess metric fidelity, complemented by an open-source benchmark repository encompassing algorithms, datasets, and evaluation tools. (3) We outline key open challenges, offering forward-looking perspectives on dynamic thresholding, model interpretability, zero-shot transfer, multi-modal alignment, and foundational models for time series. By addressing existing theoretical and empirical limitations, this survey establishes a coherent framework to guide future TSAD research and practical deployments.
Figures & tables
Fig. 1 : Publication trends of TSAD algorithms from 2015 to 2025. The top panel shows a statistical breakdown based on the taxonomy of algorithmic paradigms, while the bottom panel delineates publications across different model architectures.
Fig. 2 : The organizational framework of this paper.
Citation
Application
Taxonomy of Paradigms
Anomaly Taxonomy
Novel Algorithms
Library
Datasets
Metrics
Discussion
Prot/SSM/Diff
LLM
Interpretability
Multi-Modality
Multi-Domain
Choi et al. [11] (2021)
✓
✓
✓
Schmidl et al. [64] (2022)
✓
✓
✓
✓
Wen et al. [76] (2023)
✓
Yan et al. [81] (2023)
✓
✓
✓
Ye et al. [85] (2024)
✓
✓
✓
✓
✓
TABLE I : Overview of Recent Key Time Series Anomaly Detection Surveys.
Fig. 3 : Illustration of different types of anomalies. The red box indicates the area of anomaly.
TABLE II : Summary of algorithmic innovations. Detailed descriptions and references are available in SM .
Fig. 5 : The mapping relationship between algorithm architectures (categorized by model capability) and the taxonomy of paradigms.
Architecture
Advantages
Disadvantages
Unified TSAD Model Exploring LLM-Powered Model Predictive LLM-based Model
①High representational capacity providing a versatile framework for cross-domain and multi-task learning. ②Seamless integration with heterogeneous modalities (such as vision and text) via multimodal LLMs for context-rich reasoning. ③Facilitates collaborative multi-agent architectures that leverage generative capabilities for automated diagnostic reasoning.
①Substantial parameter footprint, demanding memory overhead, and elevated inference latency. ②Evolving research domain without consistent empirical margins over specialized architectures. ③Heavy reliance on textual tokenization or 2D image rendering, causing token bloat in high-frequency numerical time series.
Diffusion-based Model
①Exceptional generative capacity for accurately capturing complex high-dimensional distributions. ②High modeling flexibility with minimal parametric assumptions on data distributions.
①Elevated computational overhead from iterative reverse denoising steps during both training and inference, limiting real-time utility. ②High data dependency requiring substantial training volumes to faithfully learn normal temporal baselines.
Transformer-based Model
①Highly effective at modeling non-local associations and pairwise variable relationships across extended sequences. ②Fully parallelized training over entire temporal sequences, accelerating model convergence.
①Quadratic computational and memory complexity with respect to sequence length, creating scalability bottlenecks.
State Space Model
①Linear computational complexity with sequence length, enabling efficient long-horizon temporal processing. ②Favorable efficiency-performance trade-off compared to dense self-attention under equivalent parameter scales.
①Relatively nascent architectural ecosystem with limited off-the-shelf tooling and pre-trained baselines. ②Empirical advantages over long horizons require careful recurrent state parameterization and domain tuning.
Recurrent Neural Network Model Graph Neural Network Model Temporal Convolution Network Model
①Intrinsic inductive biases tailored for causal sequential dependencies. ②Moderate computational footprint suitable for deployment under tight edge resources. ③Explicit structural modeling of inter-variable topological graphs when paired with graph convolutions.
①Susceptible to gradient vanishing and memory bottlenecks over long temporal windows. ②Constrained transferability compared to attention-based architectures, limiting cross-domain generalization.
Prototype-based Model
①Explicitly characterizes normal manifolds using compact sets of learnable prototypes. ②Modular framework easily integrated into diverse metric and representation backbones.
①Prototype representations require careful regularization to ensure geometric diversity and prevent representation collapse. ②Scaling to large numbers of prototypes substantially increases optimization difficulty and inference latency.
TABLE III : Comparative analysis of strengths and limitations across representation architectures.
TABLE IV : The algorithm selection suggestions under different selection criteria.
Figure 10Figure 11Figure 12Figure 13
Qiaoli Ni is with HUAWEI Technologies Co., Ltd., Shenzhen, Guangdong 518066, China.
Time-Series Anomaly Detection identifies unusual patterns or events in sequential data, from sensor readings and network traffic to health and industrial telemetry.