The DOI matches its Crossref record; 19 of the 20 references it cites were found in a registry or an index.
USAD: UnSupervised Anomaly Detection on Multivariate Time Series
Open accessAbstract
The automatic supervision of IT systems is a current challenge at Orange. Given the size and complexity reached by its IT operations, the number of sensors needed to obtain measurements over time, used to infer normal and abnormal behaviors, has increased dramatically making traditional expert-based supervision methods slow or prone to errors. In this paper, we propose a fast and stable method called UnSupervised Anomaly Detection for multivariate time series (USAD) based on adversely trained autoencoders. Its autoencoder architecture makes it capable of learning in an unsupervised way. The use of adversarial training and its architecture allows it to isolate anomalies while providing fast training. We study the properties of our methods through experiments on five public datasets, thus demonstrating its robustness, training speed and high anomaly detection performance. Through a feasibility study using Orange's proprietary data we have been able to validate Orange's requirements on scalability, stability, robustness, training speed and high performance.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“It is beneficial to introduce an adversarial-based SSL, which can further amplify the difference between normal and anomalous data, such as USAD [66] and DAEMON [184].”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [66]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The core idea behind USAD is to amplify the reconstruction error by adversarial training between two BAEs.”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [66]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“USAD [66] is a time series anomaly detection framework that includes two BAE models, and two BAE are defined as AE 1 and AE 2 , respectively.”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [66]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Adversarial-based regularization constraints can amplify the score difference between anomalies and normal points, thereby enhancing the accuracy of anomalies identification from another perspective, such as Minmax learning [66], [163] and adversarial learning [184].”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [66]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“RANSynCoders [63], USAD[66], AnomalyTrans [163], and DAEMON [184] employ autoencoder-based reconstruction SSL framework.”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [66]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 1×
“poral data, methods based on recurrent neural networks [7] (RNNs) are very popular.”
MethodGAN(Generative Adversarial Nets)
(2017) · Journal of Japan Society for Fuzzy Theory and Intelligent Informatics · doi:10.3156/jsoft.29.5_177_2 · Cited by 19,270
Learning Internal Representations by Error Propagation
Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1985) · doi:10.21236/ada164453 · Cited by 16,172
Liu, F. T., Ting, K. M., & Zhou, Z.-H. (2008) · 2008 Eighth IEEE International Conference on Data Mining · doi:10.1109/icdm.2008.17 · Cited by 6,601 · cited 1×
“Figure 2(C) presents the results for m ∈ [ 5 , 10 , 20 , 40 , 100 ] . dimension for Z causes a large loss of information at the encoding .”
Background5th International Conference on Learning Representations (ICLR 17)
Havrylov, S., & Titov, I. (2017) · Cited by 3,503
Robust Anomaly Detection for Multivariate Time Series through Stochastic Recurrent Neural Network
Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., & Pei, D. (2019) · Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3292500.3330672 · Cited by 1,804 · cited 9×
“On top, the results obtained with SWaT and WADI datasets are presented, whereas the bottom part of the table reports obtained results from the benchmark proposed by [17], using three remaining datasets.”
BackgroundMethodResultInfluentialDetecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding
Hundman, K., Constantinou, V., Laporte, C., Colwell, I., & Soderstrom, T. (2018) · Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3219819.3219845 · Cited by 1,771
Schwartz, R., Dodge, J., Smith, N. A., & Etzioni, O. (2020) · Communications of the ACM · doi:10.1145/3381831 · Cited by 1,647
Deep Learning for Anomaly Detection: A Survey
Chalapathy, R., & Chawla, S. (2019) · arXiv · doi:10.48550/arxiv.1901.03407 · Cited by 1,240 · cited 1×
“Detecting unexpected behavior on a set of measurements correlated with each other over time is an active research discipline called anomaly detection in multivariate time series [2].”
BackgroundA Multimodal Anomaly Detector for Robot-Assisted Feeding Using an LSTM-Based Variational Autoencoder
Park, D., Hoshi, Y., & Kemp, C. C. (2018) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2018.2801475 · Cited by 1,115
Research works printed but not linked: 3
- Ian J. Goodfellow , Jean Pouget-Abadie , Mehdi Mirza , Bing Xu , David Warde-Farley , Sherjil Ozair , Aaron C. Courville , and Yoshua Bengio . 2014 . Generative Adversarial Nets. In Advances in Neural Information Processing Systems 27: Annual Conference on Neural Information Processing Systems 2014. Montreal, Quebec, Canada, 2672--2680. Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron C. Courville, and Yoshua Bengio. 2014. Generative Adversarial Nets. In Advances in Neural Information Processing Systems 27: Annual Conference on Neural Information Processing Systems 2014. Montreal, Quebec, Canada, 2672--2680.
- Chuxu Zhang , Dongjin Song , Yuncong Chen , Xinyang Feng , Cristian Lumezanu , Wei Cheng , Jingchao Ni , Bo Zong , Haifeng Chen , and Nitesh V. Chawla . 2019. A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data . Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 33 ( 2019 ), 1409--1416. Chuxu Zhang, Dongjin Song, Yuncong Chen, Xinyang Feng, Cristian Lumezanu, Wei Cheng, Jingchao Ni, Bo Zong, Haifeng Chen, and Nitesh V. Chawla. 2019. A Deep Neural Network for Unsupervised Anomaly Detection and Diagnosis in Multivariate Time Series Data. Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 33 (2019), 1409--1416.
- Bo Zong , Qi Song , Martin Renqiang Min , Wei Cheng , Cristian Lumezanu , Daeki Cho , and Haifeng Chen . 2018 . Deep autoencoding Gaussian mixture model for unsupervised anomaly detection . In 6th International Conference on Learning Representations, ICLR 2018. Toulon, France, 1--19. Bo Zong, Qi Song, Martin Renqiang Min, Wei Cheng, Cristian Lumezanu, Daeki Cho, and Haifeng Chen. 2018. Deep autoencoding Gaussian mixture model for unsupervised anomaly detection. In 6th International Conference on Learning Representations, ICLR 2018. Toulon, France, 1--19.
Cited by
TranAD: deep transformer networks for anomaly detection in multivariate time series data
Tuli, S., Casale, G., & Jennings, N. R. (2022) · Proceedings of the VLDB Endowment · doi:10.14778/3514061.3514067 · Cited by 1,063
Deep Learning for Time Series Anomaly Detection: A Survey
Zamanzadeh Darban, Z., Webb, G. I., Pan, S., Aggarwal, C., & Salehi, M. (2025) · ACM Computing Surveys · doi:10.1145/3691338 · Cited by 613
“Audibert et al. [10] propose Unsupervised Anomaly Detection (USAD) using AEs in which adversarially trained AEs are utilised to amplify reconstruction errors in MTS, distinguishing anomalies and facilitating quick learning.”
InfluentialDeep Learning for Anomaly Detection in Time-Series Data: Review, Analysis, and Guidelines
Choi, K., Yi, J., Park, C., & Yoon, S. (2021) · IEEE Access · doi:10.1109/access.2021.3107975 · Cited by 588
“Audibert et al. [72] proposed AE, in which one encoder and two decoders are trained adversarially, to identify network anomalies.”
BackgroundMethodInfluentialLi, Z., Zhao, Y., Han, J., Su, Y., Jiao, R., Wen, X., & Pei, D. (2021) · Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining · doi:10.1145/3447548.3467075 · Cited by 464
“Thus, we adopt the point-adjust approach, which is proposed by [33] and widely used in the evaluation of detection tasks [2, 23, 27, 30].”
BackgroundMethodInfluentialPractical Approach to Asynchronous Multivariate Time Series Anomaly Detection and Localization
Abdulaal, A., Liu, Z., & Lancewicki, T. (2021) · Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining · doi:10.1145/3447548.3467174 · Cited by 423
“Alternatively, USAD is based on adversarial learning.”
BackgroundMethodResultInfluentialMST-GAT: A multimodal spatial–temporal graph attention network for time series anomaly detection
Ding, C., Sun, S., & Zhao, J. (2023) · Information Fusion · doi:10.1016/j.inffus.2022.08.011 · Cited by 309
An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series
Garg, A., Zhang, W., Samaran, J., Savitha, R., & Foo, C.-S. (2022) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2021.3105827 · Cited by 285
“score even when a large number of events are not detected [8].”
BackgroundMethodResultInfluentialSelf-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects
Zhang, K., Wen, Q., Zhang, C., Cai, R., Jin, M., Liu, Y., Zhang, J. Y., Liang, Y., Pang, G., Song, D., et al. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3387317 · Cited by 295
Towards a Rigorous Evaluation of Time-Series Anomaly Detection
Kim, S., Choi, K., Choi, H.-S., Lee, B., & Yoon, S. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i7.20680 · Cited by 194
“Figure 4 plots F1 PA and F1 for SWaT and WADI, as re-ported by the original studies on USAD, DAGMM, LSTM-VAE, OmniAnomaly, and GDN.”
BackgroundMethodInfluentialYan, P., Abdulkadir, A., Luley, P.-P., Rosenthal, M., Schatte, G. A., Grewe, B. F., & Stadelmann, T. (2024) · IEEE Access · doi:10.1109/access.2023.3349132 · Cited by 212
“Audibert et al. propose a fast and stable method – Unsupervised anomaly detection for multivariate time series [92], based on adversely trained autoencoders.”
MethodRTFN: A robust temporal feature network for time series classification
Xiao, Z., Xu, X., Xing, H., Luo, S., Dai, P., & Zhan, D. (2021) · Information Sciences · doi:10.1016/j.ins.2021.04.053 · Cited by 184
Calibrated One-Class Classification for Unsupervised Time Series Anomaly Detection
Xu, H., Wang, Y., Jian, S., Liao, Q., Wang, Y., & Pang, G. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3393996 · Cited by 162
“Therefore, the vast majority of previous studies [2], [3], [4], [5], [7], [10], [11] employ point-adjust strategy in their experiment protocols.”
BackgroundMethodInfluentialEadro: An End-to-End Troubleshooting Framework for Microservices on Multi-source Data
Lee, C., Yang, T., Chen, Z., Su, Y., & Lyu, M. R. (2023) · 2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE) · doi:10.1109/icse48619.2023.00150 · Cited by 149
TFAD: A Decomposition Time Series Anomaly Detection Architecture with Time-Frequency Analysis
Zhang, C., Zhou, T., Wen, Q., & Sun, L. (2022) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3511808.3557470 · Cited by 209
ImDiffusion: Imputed Diffusion Models for Multivariate Time Series Anomaly Detection
Chen, Y., Zhang, C., Ma, M., Liu, Y., Ding, R., Li, B., He, S., Rajmohan, S., Lin, Q., & Zhang, D. (2023) · Proceedings of the VLDB Endowment · doi:10.14778/3632093.3632101 · Cited by 139
Learning Sparse Latent Graph Representations for Anomaly Detection in Multivariate Time Series
Han, S., & Woo, S. S. (2022) · Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining · doi:10.1145/3534678.3539117 · Cited by 137
“Considering most cases where the trade-off between FPs and FNs is inevitable, in practice, maintenance technicians with domain expertise will prefer a high level of sensitivity over specificity to avoid missing any critical incidents that are worthwhile inspecting for future reference [5].”
BackgroundA survey of AIOps methods for failure management
Notaro, P., Cardoso, J., & Gerndt, M. (2021) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3483424 · Cited by 142
DTAAD: Dual Tcn-attention networks for anomaly detection in multivariate time series data
Yu, L., Lu, Q., & Xue, Y. (2024) · Knowledge-Based Systems · doi:10.1016/j.knosys.2024.111849 · Cited by 145
Yu, G., Chen, P., Li, Y., Chen, H., Li, X., & Zheng, Z. (2023) · Proceedings of the 31st ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering · doi:10.1145/3611643.3616249 · Cited by 140
“, USAD [2] and JumpStarter [43]) and monitor additional metrics beyond the P90 latency and success ratio to avoid missing faults.”
BackgroundInfluentialTime Series Anomaly Detection With Adversarial Reconstruction Networks
Liu, S., Zhou, B., Ding, Q., Hooi, B., Zhang, Z., Shen, H., & Cheng, X. (2022) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2021.3140058 · Cited by 152
“The USAD [25] utilizes adversely trained autoencoders to learn data features in an unsupervised way while achieving a high performance.”
MethodImputation-based Time-Series Anomaly Detection with Conditional Weight-Incremental Diffusion Models
Xiao, C., Gou, Z., Tai, W., Zhang, K., & Zhou, F. (2023) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3580305.3599391 · Cited by 130
“USAD [3] presents an autoencoder architecture whose adversarial-style learning is also inspired by GAN.”
BackgroundCorrelation-Aware Spatial–Temporal Graph Learning for Multivariate Time-Series Anomaly Detection
Zheng, Y., Koh, H. Y., Jin, M., Chi, L., Phan, K. T., Pan, S., Chen, Y. P., & Xiang, W. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3325667 · Cited by 123
“The six state-of-the-art frameworks for multivariate time-series anomaly detection are LSTM-VAE [26], Omni-Anomaly [3], UnSupervised Anomaly Detection on multi-variate time series (USAD) [68], multivariate time-series anomaly detection via GAT (MTAD-GAT) [63], GDN [11], and InterFusion [1].”
MethodInfluentialFeng, C., & Tian, P. (2021) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3447548.3467137 · Cited by 119
“USAD comes from a PyTorch implementation in the Github repository5.”
BackgroundMethodWang, F., Jiang, Y., Zhang, R., Wei, A., Xie, J., & Pang, X. (2025) · Sensors · doi:10.3390/s25010190 · Cited by 119
“To address this issue, USAD [39] was proposed, utilizing adversarial training to enhance the stability and speed of autoencoder-based anomaly detection.”
Zeng, G., Yang, Y.-W., Lu, K., Geng, G., & Weng, J. (2025) · IEEE Transactions on Reliability · doi:10.1109/tr.2025.3528256 · Cited by 120
Song, Y., Xin, R., Chen, P., Zhang, R., Chen, J., & Zhao, Z. (2023) · Future Generation Computer Systems · doi:10.1016/j.future.2023.03.020 · Cited by 100
Robust anomaly detection for multivariate time series through temporal GCNs and attention-based VAE
Shi, Y., Wang, B., Yu, Y., Tang, X., Huang, C., & Dong, J. (2023) · Knowledge-Based Systems · doi:10.1016/j.knosys.2023.110725 · Cited by 102
GAN-Based Anomaly Detection for Multivariate Time Series Using Polluted Training Set
Du, B., Sun, X., Ye, J., Cheng, K., Wang, J., & Sun, L. (2021) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2021.3128667 · Cited by 98
“In order to compare the performance of our model with the one proposed by Audibert et al. [30], we also assess performance on SWaT and WADI without point-adjust .”
MethodResultInfluentialZhang, Q., Xu, C., Li, J., Sun, Y., Bao, J., & Zhang, D. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.125861 · Cited by 119
Zeng, F., Chen, M., Qian, C., Wang, Y., Zhou, Y., & Tang, W. (2023) · Future Generation Computer Systems · doi:10.1016/j.future.2023.02.015 · Cited by 103
Anomaly Detection in Connected and Autonomous Vehicles: A Survey, Analysis, and Research Challenges
Baccari, S., Hadded, M., Ghazzai, H., Touati, H., & Elhadef, M. (2024) · IEEE Access · doi:10.1109/access.2024.3361829 · Cited by 108
“However, these models suffer from a higher number of false positives/negatives, since they do not have an annotated schema to differentiate clearly between normal and abnormal data [73], [74].”
BackgroundMejri, N., Lopez-Fuentes, L., Roy, K., Chernakov, P., Ghorbel, E., & Aouada, D. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.124922 · Cited by 133
“Given the recent advances in DL, DNNs have been considered as an alternative (Audibert et al., 2020; Li et al., 2019; Malhotra et al., 2015; Su et al., 2019), mainly taking inspiration from traditional methods.”
InfluentialHao, J., Chen, P., Chen, J., & Li, X. (2024) · Future Generation Computer Systems · doi:10.1016/j.future.2024.05.006 · Cited by 75
Robust Multimodal Failure Detection for Microservice Systems
Zhao, C., Ma, M., Zhong, Z., Zhang, S., Tan, Z., Xiong, X., Yu, L., Feng, J., Sun, Y., Zhang, Y., et al. (2023) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3580305.3599902 · Cited by 112
“We apply five popular single-modal anomaly detection meth-ods (Section 5.1), JumpStarter [25], USAD [1], LogAnomaly [27], Deeplog [5], TraceAnomaly [21], and the combination of Jump-Starter,LogAnomaly,andTraceAnomaly,toconductmetric/log/trace anomalydetection,respectively.”
BackgroundMethodInfluentialUnsupervised Deep Learning for IoT Time Series
Liu, Y., Zhou, Y., Yang, K., & Wang, X. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3243391 · Cited by 81
“[127] performed adversarial training on two AEs.”
Heterogeneous Anomaly Detection for Software Systems via Semi-supervised Cross-modal Attention
Lee, C., Yang, T., Chen, Z., Su, Y., Yang, Y., & Lyu, M. R. (2023) · 2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE) · doi:10.1109/icse48619.2023.00148 · Cited by 68
Wang, C., & Liu, G. (2024) · Advanced Engineering Informatics · doi:10.1016/j.aei.2024.102357 · Cited by 76
Xin, R., Liu, H., Chen, P., & Zhao, Z. (2023) · Journal of Cloud Computing Advances Systems and Applications · doi:10.1186/s13677-022-00383-6 · Cited by 77
“[7] developed the USAD based on an adversely trained AutoEncoder and achieved the best detection accuracy.”
MethodGraph Structure Change-Based Anomaly Detection in Multivariate Time Series of Industrial Processes
Zhang, Z., Geng, Z., & Han, Y. M. (2024) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2023.3347000 · Cited by 66
PreGAN: Preemptive Migration Prediction Network for Proactive Fault-Tolerant Edge Computing
Tuli, S., Casale, G., & Jennings, N. R. (2022) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom48880.2022.9796778 · Cited by 82
“To model the dependence of a data point x t at a timestamp t , we consider the sliding window of length k , denoted as as is done in prior work [32].”
MethodHe, H., Li, X., Chen, P., Chen, J., Liu, M., & Wu, L. (2024) · Journal of Cloud Computing · doi:10.1186/s13677-024-00677-x · Cited by 62
Huang, T., Chen, P., & Li, R. (2022) · Proceedings of the ACM Web Conference 2022 · doi:10.1145/3485447.3511984 · Cited by 61
Zhang, Z., Li, W., Ding, W., Zhang, L., Lu, Q., Hu, P., Gui, T., & Lu, S. (2022) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3572780 · Cited by 70
“We apply the synthetic dataset on three state-of-the-art deep learning-based anomaly detection methods (LSTM-VAE [23], MAD-GAN [20], and USAD [6]) to evaluate the performance (in terms of F1-score) of anomaly detection with different ratio of injected noises, whose results are shown in Figure 1(b).”
BackgroundMethodInfluentialTushkanova, O. N., Levshun, D., Branitskiy, A. A., Владимировна, Ф. Е., Novikova, E. S., & Kotenko, I. V. (2023) · Algorithms · doi:10.3390/a16020085 · Cited by 71
A Review of Neural Networks for Anomaly Detection
Filho, J. E. D. A., Brandão, L. C. P., Fernandes, B., & Maciel, A. M. A. (2022) · IEEE Access · doi:10.1109/access.2022.3216007 · Cited by 88
“architecture of the autoencoder made it able to learn unsupervised and using adversarial training and its architecture allows it to isolate anomalies while giving fast training [61].”
BackgroundNavigating the metric maze: a taxonomy of evaluation metrics for anomaly detection in time series
Sørbø, S., & Ruocco, M. (2024) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-023-00988-8 · Cited by 58
“The point adjusted metrics were first introduced by [30], and has been widely used in TSAD [51, 22, 52, 39, 53, 54, 36, 1, 55, 35, 56, 24, 57, 58, 59, 34].”
InfluentialVerma, P., Bharot, N., Breslin, J. G., O’Shea, D., Vidyarthi, A., & Gupta, D. K. (2023) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2023.3335385 · Cited by 65
“Audibert et al. [23] put forth an anomaly detection approach in an unsupervised way and leveraging AEs for multivariate time series.”
BackgroundLearning Robust Deep State Space for Unsupervised Anomaly Detection in Contaminated Time-Series
Li, L., Yan, J., Wen, Q., Jin, Y., & Yang, X. (2022) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2022.3171562 · Cited by 57
“Models such as Omnianomaly [10] and USAD [11], although assuming clean training data, can still be found experimentally to remain robust in noisy training data due to the implicit robustness contained in the probabilistic models.”
BackgroundMethodInfluentialSpacecraft anomaly detection with attention temporal convolution networks
Liu, L., Tian, L., Kang, Z., & Wan, T. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-08213-9 · Cited by 62
“They use some popular architectures, including long short-term memory networks (LSTMs) [18,19], autoencoders (AE) [20,21,22], generative adversarial networks (GANs) [23,24,25], and Transformer [26,27].”
BackgroundHybrid Anomaly Detection via Multihead Dynamic Graph Attention Networks for Multivariate Time Series
Zhou, L., Zeng, Q., & Li, B. (2022) · IEEE Access · doi:10.1109/access.2022.3167640 · Cited by 69
“USAD is an unsupervised method based on reconstruction and consists of three parts, an encoder and 2 decoders that share the same encoder network.”
MethodInfluentialMultiscale Wavelet Graph AutoEncoder for Multivariate Time-Series Anomaly Detection
Wang, J., Shao, S., Bai, Y., Deng, J., & Lin, Y. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3223142 · Cited by 56
“That is probably because AEs can also reconstruct abnormal patterns well only learning from normal distribution [10] due to their powerful generalization ability.”
BackgroundMethodInfluentialQi, S., Chen, J., Chen, P., Wen, P., Niu, X., & Xu, L. (2023) · The Journal of Supercomputing · doi:10.1007/s11227-023-05534-3 · Cited by 55
Stgat-Mad : Spatial-Temporal Graph Attention Network For Multivariate Time Series Anomaly Detection
Zhan, J., Wang, S., Ma, X., Wu, C., Yang, C., Zeng, D., & Wang, S. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9747274 · Cited by 55
“…0.915 0.409 0.570 0.748 0.990 0.490 0.658 0.931 0.990 0.725 0.821 0.757 LSTM-VAE [13] 0.766 0.979 0.860 0.878 0.910 0.603 0.720 0.800 0.990 0.669 0.802 0.849 0.990 0.667 0.790 0.720 USAD[15] 0.960 0.347 0.510 0.755 0.834 0.159 0.267 0.647 0.979 0.447 0.615 0.908 0.990 0.484 0.652 0.565 MTAD-GAT[19]”
BackgroundMethodResultInfluentialTemporal-Frequency Masked Autoencoders for Time Series Anomaly Detection
Fang, Y., Xie, J., Zhao, Y. L., Lü, C., Gao, Y., & Zheng, K. (2024) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde60146.2024.00099 · Cited by 110
“This comparison substantiates that our contrastive criterion can effectively mitigate time series distribution shifts, resulting in higher generalization in the threshold.”
MethodResultPatchAD: A Lightweight Patch-Based MLP-Mixer for Time Series Anomaly Detection
Zhong, Z., Yu, Z., Yang, Y., Wang, W. W., Yang, K., & Chen, C. L. P. (2025) · IEEE Transactions on Big Data · doi:10.1109/tbdata.2025.3596745 · Cited by 89
“Similarly, Yang et al. [11] extended the paradigm of image contrastive learning established by BYOL [20] by designing different temporal views to circumvent the challenges associated with negative sample construction in contrastive learning.”
Efficient KPI Anomaly Detection Through Transfer Learning for Large-Scale Web Services
Zhang, S., Zhong, Z., Li, D., Fan, Q., Sun, Y., Zhu, M., Zhang, Y., Pei, D., Sun, J., Liu, Y., et al. (2022) · IEEE Journal on Selected Areas in Communications · doi:10.1109/jsac.2022.3180785 · Cited by 51
“Although MTS can give a more comprehensive picture of the monitored entity and MTS anomaly detection has gained much attention recently [9], [10], [14]–[18], detecting anomalies in UTS is still essential in modern operation tasks.”
BackgroundMethodExplainable time series anomaly detection using masked latent generative modeling
Lee, D.-S., Malacarne, S., & Aune, E. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.110826 · Cited by 61
Detection of Anomaly using Machine Learning: A Comprehensive Survey
Mane, D., Sangve, S. M., Upadhye, G. D., Kandhare, S., Mohole, S., Sonar, S., & Tupare, S. (2022) · International Journal of Emerging Technology and Advanced Engineering · doi:10.46338/ijetae1122_15 · Cited by 52
Xie, T., Xu, Q., & Jiang, C. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.120725 · Cited by 50
Twin Graph-Based Anomaly Detection via Attentive Multi-Modal Learning for Microservice System
Huang, J., Yang, Y., Yu, H., Li, J., & Zheng, X. (2023) · 2023 38th IEEE/ACM International Conference on Automated Software Engineering (ASE) · doi:10.1109/ase56229.2023.00138 · Cited by 50
“Before delving into the details of our model, we first introduce different attention mechanisms used in the following sections.”
MethodLabel-Free Multivariate Time Series Anomaly Detection
Zhou, Q., He, S., Liu, H., Chen, J., & Meng, W. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3349613 · Cited by 67
“USAD [43] and DAEMON [8] use adversarial learning to promote reconstruction quality.”
BackgroundMethodGraph Anomaly Detection in Time Series: A Survey
Ho, T. K. K., Karami, A., & Armanfard, N. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3566620 · Cited by 89
“Most existing TSAD algorithms are designed to analyze time-series data as a univariate modality [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17] or as multiple individual modalities (variables) [18], [19], [20], [21], [21], [22], [23], [24].”
Sarkon, G. K., Safaei, B., Kenevisi, M. S., Arman, S., & Zeeshan, Q. (2022) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-022-09786-9 · Cited by 49
ADTCD: An Adaptive Anomaly Detection Approach Toward Concept Drift in IoT
Xu, L., Xiao, D., Peng, H., Zhao, D., & Li, X. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3265964 · Cited by 47
“2) Compared Algorithms: We compare ADTCD with streaming variants of three state-of-the-art anomaly detection approaches based on RNNs, including DAGMM [16], REBM [17] and EncDec-AD, an offline anomaly detection algorithms USAD [25], and three flow anomaly detection algorithm ARCUS [20], MemStream [26], and OGMMFVRD [2…”
MethodInfluentialAn Adversarial Time–Frequency Reconstruction Network for Unsupervised Anomaly Detection
Jin, F., Wang, Z., Wu, H., Sun, D., Wu, J., & Lu, X. B. (2023) · Neural Networks · doi:10.1016/j.neunet.2023.09.018 · Cited by 46
SDFVAE: Static and Dynamic Factorized VAE for Anomaly Detection of Multivariate CDN KPIs
Dai, L., Lin, T., Liu, C., Jiang, B., Liu, Y., Xu, Z., & Zhang, Z.-L. (2021) · International World Wide Web Conference (WWW) · doi:10.1145/3442381.3450013 · Cited by 50
“large number of individual models for each metric, but also helps increase the effectiveness of anomaly detection, as an incident typically tends to produce anomalies in multiple KPIs [2, 21, 31, 39].”
BackgroundContrastive Time-Series Anomaly Detection
Kim, H., Kim, S., Min, S., & Lee, B. (2023) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2023.3335317 · Cited by 64
“For the comparison with classical methods and existing time-series anomaly detection methods, the results in [40] were referred to.”
BackgroundMethodResultInfluentialLi, Z., Zhao, Y., Geng, Y., Zhao, Z., Wang, H., Chen, W., Jiang, H., Vaidya, A., Su, L., & Pei, D. (2022) · IEEE Journal on Selected Areas in Communications · doi:10.1109/jsac.2022.3191341 · Cited by 45
“Moreover, following previous works [2], [3], [13], [15] in MTS anomaly detection, we enumerate and find the optimal threshold to convert anomaly scores given by detection model D to detection set P , and then calculate the best range-based F1-score.”
BackgroundMethodInfluentialHe, X., Li, Y., Tan, J. R., Wu, B., & Li, F. (2023) · Proceedings of the VLDB Endowment · doi:10.14778/3583140.3583155 · Cited by 46
Time Series Anomaly Detection using Diffusion-based Models
Pintilie, I., Manolache, A., & Brad, F. (2023) · IEEE ... International Conference on Data Mining workshops · doi:10.1109/icdmw60847.2023.00080 · Cited by 63
ICS anomaly detection based on sensor patterns and actuator rules in spatiotemporal dependency
Cai, J., Wei, Z., & Luo, J. (2024) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2024.3393528 · Cited by 38
“There are five state-of-the-art anomaly detection methods, in-cludingUSAD[11],LSTM-VAE[14],GDN[19],GTA[21],and MTAD-GAT[24].”
BackgroundMethodAnomaly VAE-Transformer: A Deep Learning Approach for Anomaly Detection in Decentralized Finance
Song, A., Seo, E., & Kim, H. (2023) · IEEE Access · doi:10.1109/access.2023.3313448 · Cited by 51
“Autoencoder (AE) [40], [48], [49], VAE [8], [9], [10], [11], [12], [50], and transformer based models [5], [6], [7], [51], [52] use such reconstruction errors.”
Sensing anomaly of photovoltaic systems with sequential conditional variational autoencoder
Li, D., Zhang, Y., Yang, Z., Jin, Y., & Xu, Y. (2023) · Applied Energy · doi:10.1016/j.apenergy.2023.122124 · Cited by 40
Adaptive Memory Broad Learning System for Unsupervised Time Series Anomaly Detection
Zhong, Z., Yu, Z., Fan, Z., Chen, C. L. P., & Yang, K. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3415621 · Cited by 40
“The presence of anomalous data within training sets often challenges algorithms [10].”
InfluentialMultiview Graph Contrastive Learning for Multivariate Time-Series Anomaly Detection in IoT
Qin, S., Chen, L., Luo, Y., & Tao, G. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3303946 · Cited by 37
“8) USAD [32]: A CNN-based model with adversarial training strategy.”
MethodSong, P., Zhao, C., Huang, B., & Ding, J. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3262277 · Cited by 37
“If the reconstruction ability is too strong, the model tends to learn an identity mapping to reconstruct input data well without extracting effective representations [18], [19].”
BackgroundCoupled Attention Networks for Multivariate Time Series Anomaly Detection
Xia, F., Chen, X., Yu, S., Hou, M., Liu, M., & You, L. (2023) · IEEE Transactions on Emerging Topics in Computing · doi:10.1109/tetc.2023.3280577 · Cited by 40
“Many researchers have started using multivariate time series data as input to improve anomaly detection accuracy [9], [11], [12], [19].”
BackgroundWang, Y., Zhou, Y., Ma, J., & Jin, Q. (2023) · Applied Soft Computing · doi:10.1016/j.asoc.2023.111169 · Cited by 36
Ye, M., Zhang, Q., Xue, X., Wang, Y., Jiang, Q., & Qiu, H. (2024) · IEEE Systems Journal · doi:10.1109/jsyst.2023.3347435 · Cited by 43
“Audibert et al. [27] proposed a two-phase training strategy to detect anomalies, where the original input is first reconstructed through an autoencoder, which is then used to generate negative samples to fool the discriminator.”
MethodLUAD: A lightweight unsupervised anomaly detection scheme for multivariate time series data
Fan, J., Liu, Z., Wu, H., Wu, J. X., Si, Z., Peng, H., & Luan, T. H. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.126644 · Cited by 36
DroneSSL: Self-Supervised Multimodal Anomaly Detection in Internet of Drone Things
Akram, J., Anaissi, A., Othman, W., Alabdulatif, A., & Akram, A. (2024) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2024.3376440 · Cited by 39
“In their work, [31] crafted a model predicated on a two-phase adversarial training approach.”
Mattera, G., Vozza, M., Polden, J., Nele, L., & Pan, Z. S. (2024) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-024-02507-y · Cited by 35
“Another issue related to what has been mentioned earlier is that when compressing a signal into the latent space, there will be some low-dimensional features that are very similar for both anomalies and normal signals (Audibert et al., 2020).”
Anomaly detection using invariant rules in Industrial Control Systems
Zhu, Q., Ding, Y., Jiang, J. S., & Yang, S. (2024) · Control Engineering Practice · doi:10.1016/j.conengprac.2024.106164 · Cited by 49
Deep Anomaly Detection on Tennessee Eastman Process Data
Hartung, F., Franks, B. J., Michels, T., Wagner, D., Liznerski, P., Reithermann, S., Fellenz, S., Jirasek, F., Rudolph, M., Neider, D., et al. (2023) · Chemie Ingenieur Technik · doi:10.1002/cite.202200238 · Cited by 40
Enhancing anomaly detection in distributed power systems using autoencoder-based federated learning
Kea, K., Han, Y., & Kim, T.-K. (2023) · PLoS ONE · doi:10.1371/journal.pone.0290337 · Cited by 41
Zhang, X., Song, C., Zhao, J., Xu, Z., & Deng, X. (2023) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2023.3333921 · Cited by 37
“…compare with the following five classical and state-of-the-art anomaly detection methods: (cid:2) PCA [39]: Principal component analysis; (cid:2) AE [40]: Autoencoder; LSTM [41]: short-term memory network; (cid:2) GDN [32]: Graph deviation network; (cid:2) MTGFlow [42]: Multivariate time series…”
MethodInterpretable Failure Localization for Microservice Systems Based on Graph Autoencoder
Sun, Y., Lin, Z., Shi, B., Zhang, S., Ma, S., Jin, P., Zhong, Z., Pan, L., Guo, Y., & Pei, D. (2024) · ACM Transactions on Software Engineering and Methodology · doi:10.1145/3695999 · Cited by 59
An explainable unsupervised anomaly detection framework for Industrial Internet of Things
Abudurexiti, Y., Han, G., Zhang, F., & Liu, L. (2024) · Computers & Security · doi:10.1016/j.cose.2024.104130 · Cited by 42
DACAD: Domain Adaptation Contrastive Learning for Anomaly Detection in Multivariate Time Series
Darban, Z. Z., Yang, Y., Webb, G. I., Aggarwal, C. C., Wen, Q., Pan, S., & Salehi, M. (2025) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2025.3569909 · Cited by 60
“However, models like USAD [40] face challenges with long time series data due to error accumulation in decoding.”
MAD-SGCN: Multivariate Anomaly Detection with Self-learning Graph Convolutional Networks
Qi, P., Li, D., & Ng, S.-K. (2022) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde53745.2022.00097 · Cited by 33
Memory-Augmented U-Transformer For Multivariate Time Series Anomaly Detection
Qin, S., Luo, Y., & Tao, G. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10096179 · Cited by 35
“We compare our model with 7 state-of-the-art baselines, including LSTM-VAE [9], OmniAnomaly [10], MTAD-GAT [21], USAD [11], TransFram [22], TranAD [15], and AnomalyTran [14].”
MethodKnowledge-Enhanced Spatiotemporal Analysis for Anomaly Detection in Process Manufacturing
Allen, L., Lu, H., & Cordiner, J. (2024) · Computers in Industry · doi:10.1016/j.compind.2024.104111 · Cited by 37
Adversarial Transformer-Based Anomaly Detection for Multivariate Time Series
Yu, X., Zhang, K., Liu, Y., Zou, B., Wang, J., Wang, W., & Qian, R. (2024) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2024.3507211 · Cited by 35
“Regrettably, GAN is prone toproblems,suchaspatterncollapseandnonconvergenceduring thetrainingprocess[11].”
InfluentialTime Series Anomaly Detection Using Transformer-Based GAN With Two-Step Masking
Shin, A.-H., Kim, S. T., & Park, G.-M. (2023) · IEEE Access · doi:10.1109/access.2023.3289921 · Cited by 38
“methods [2], [8], [9], [10], [11], [12], [13], [14], [15], many studies have focused on the unsupervised learning methods due to the lack of labeled anomalies and highly nonlinear temporal dependencies.”
BackgroundAutonomous selection of the fault classification models for diagnosing microservice applications
Song, Y., Xin, R., Chen, P., Zhang, R., Chen, J., & Zhao, Z. (2023) · Future Generation Computer Systems · doi:10.1016/j.future.2023.12.005 · Cited by 33
“The eight models are as follows: WPS [37], CAE_M [38], CGNN-MHSA-AR [39], MAD_GAN [40], USAD [41], DAGMM [42], MTAD-GAT [43], OmniAnomaly [44].”
GAT-AD: Graph Attention Networks for contextual anomaly detection in network monitoring
Latif-Martínez, H., Suárez‐Varela, J., Cabellos‐Aparicio, A., & Barlet‐Ros, P. (2024) · Computers & Industrial Engineering · doi:10.1016/j.cie.2024.110830 · Cited by 44
Gao, H., Xin, R., Chen, P., Li, X., Lü, N., & You, P. (2025) · Journal of Cloud Computing Advances Systems and Applications · doi:10.1186/s13677-025-00766-5 · Cited by 37
Yahya, M. A., Moya, A. R., & Ventura, S. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11401-9 · Cited by 54
Di, Y., Wang, F., Zhao, Z., Zhai, Z., & Chen, X. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.124348 · Cited by 32
Adversarial Graph Neural Network for Multivariate Time Series Anomaly Detection
Zheng, B., Ming, L., Zeng, K., Zhou, M., Zhang, X., Ye, T., Yang, B., Zhou, X., & Jensen, C. S. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3419891 · Cited by 34
Metadata
- Authors
- Audibert, J., Michiardi, P., Guyard, F., Marti, S., & Zuluaga, M. A.
- Affiliations
- EURECOM (FR)
- EURECOM, Biot, France
- Orange (France) (FR)
- Venue
- Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining
- Publisher
- ACM
- Published
- 2020-08-20
- Pages
- 3395-3404
- OpenAlex
- W3081497074
- Cited by
- 1,094
- References
- 20
- Citations by year
- 2026: 281 · 2025: 266 · 2024: 256 · 2023: 170 · 2022: 84 · 2021: 37
- FWCI
- 27.79
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Anomaly Detection Techniques and Applications
- Physical Sciences › Computer Science › Computer Networks and Communications › Software System Performance and Reliability
- Physical Sciences › Computer Science › Computer Networks and Communications › Network Security and Intrusion Detection
- Keywords
- unsupervised anomaly detection, multivariate time series, adversarial autoencoder, network anomaly detection, orange, scalability, robustness
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.