USAD: UnSupervised Anomaly Detection on Multivariate Time Series
Open accessThe DOI matches its Crossref record; 19 of the 20 references it cites were found in a registry or an index.
Abstract
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
Items 301–400 of 1,000
Deep Learning-Based Anomaly Detection in Cyber-Physical System
Oswal, S., Shinde, S. K., & Vijayalakshmi, M. (2023) · Transactions on computer systems and networks · doi:10.1007/978-981-99-4518-4_4 · Cited by 9
Wang, Z., Zhang, L., Qiu, Q., & Kong, F. (2023) · IEEE Real-Time Systems Symposium (RTSS) · doi:10.1109/rtss59052.2023.00016 · Cited by 9
“…(VAE) [26], Convolutional Neural Network (CNN) [27], [28], Long Short-Term Memory (LSTM) [29]–[31], and Generative Adversarial Network (GAN) [32], [33], to detect sensor attacks in different CPS applications such as autonomous vehicles [3], [34], aerial systems [35], [36], smart grids [37],…”
BackgroundAnomaly Detection in Multi-Host Environment Based on Federated Hypersphere Classifier
Kwon, J., Jung, B., Lee, H., & Lee, S. (2022) · Electronics · doi:10.3390/electronics11101529 · Cited by 11
RCAD: Real-time Collaborative Anomaly Detection System for Mobile Broadband Networks
Ahmed, A. H., Riegler, M. A., Hicks, S. A., & Elmokashfi, A. M. (2022) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3534678.3539097 · Cited by 10
Towards the Testbed and Dataset for Analysis of Water Treatment Systems Security
Владимировна, Ф. Е., Novikova, E. S., Danilov, A., & Saenko, I. B. (2024) · Lecture notes in networks and systems · doi:10.1007/978-981-99-7814-4_37 · Cited by 8
Yang, S., Li, J., Shi, K., Chen, Y., Zhu, Y., He, X., Jinlong, W., & Pan, C. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10448347 · Cited by 8
“Following the convention [10, 11, 26], we use precision, recall and F1-score as the evaluation metrics.”
MethodKamalakannan, U., Ramamoorthy, S., & Ramamurthi, P. (2024) · Systems Engineering · doi:10.1002/sys.21755 · Cited by 12
Semi-Supervised Federated Analytics for Heterogeneous Household Characteristics Identification
Chen, W., Bu, S., Zhang, X., Tao, Y., Zhang, Y., & Han, Z. (2024) · IEEE Transactions on Smart Grid · doi:10.1109/tsg.2024.3415504 · Cited by 10
“When dealing with different sampling rates, the most common technique is to down-sampling the minute-level data [31] but it can have a negative effect if the fine-grained information is lost [32].”
DGTAD: decomposition GAN-based transformer for anomaly detection in multivariate time series data
Chen, Z., Yu, J., Tan, Q., Li, S., & Du, X. (2024) · Applied Intelligence · doi:10.1007/s10489-024-05693-7 · Cited by 9
Xie, T., Gao, Z., Xu, Q., Jiang, C., & Zhang, A. (2024) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2024.3485808 · Cited by 11
“To enhance reconstruction capability, [22] and [23] incorporated adversarial strategies from generative adversarial networks into autoencoders.”
MAGE: Multiperiodic Adaptive Graph Evolution Guided Anomaly Detection in Industrial IoT
Wang, W., Zhou, X., Qiu, T., & Wang, L. (2025) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2025.3558341 · Cited by 9
Kong, Y., Guo, J., Ren, L., Kang, G., Wang, Y., & Lü, J. (2025) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2025.3563540 · Cited by 8
Huang, Y., Lin, L., Wang, X., Garg, S., Moussa, S., & Alrashoud, M. R. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3577566 · Cited by 8
“Audibert et al. [36] proposed an unsupervised anomaly detection method based on adversarially trained autoencoders, capable of isolating anomalies while enabling fast training.”
Embedding Models for Multivariate Time Series Anomaly Detection in Industry 5.0
Colombi, L., Vespa, M., Belletti, N., Brina, M., Dahdal, S., Tabanelli, F., Resca, F., Bellodi, E., Tortonesi, M., Stefanelli, C., et al. (2025) · Data Science and Engineering · doi:10.1007/s41019-025-00295-w · Cited by 18
Duan, J., & Feng, J. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.111724 · Cited by 10
Deep anomaly detection of temporal heterogeneous data in AIOps: a survey
Gui, J., Ma, Z., Zhou, H., Su, Y., Zhang, M., Yu, K., & Wu, X. (2025) · Frontiers of Information Technology & Electronic Engineering · doi:10.1631/fitee.2400467 · Cited by 12
“Another example is the various generative models, such as generative adversarial network (GAN) (Li D et al., 2019b; Audibert et al., 2020) and Transformer (Tuli et al., 2022).”
Sparse and semi-attention guided faults diagnosis approach for distributed online services
Zhang, L., & Shi, Y. (2023) · Applied Soft Computing · doi:10.1016/j.asoc.2023.110911 · Cited by 8
SA2E-AD: A Stacked Attention Autoencoder for Anomaly Detection in Multivariate Time Series
Li, M., Li, Z., Yang, Z., Zhou, X., Li, Y., Wu, Z., Kong, L., & Nai, K. (2024) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3653677 · Cited by 10
Malicious Participants and Fake Task Detection Incorporating Gaussian Bias
Wang, J., Zhao, D., & Zhao, G. (2024) · ACM Transactions on Internet Technology · doi:10.1145/3696419 · Cited by 9
Wang, D., Enlund, T., Trygg, J., Tysklind, M., & Jiang, L. (2022) · IEEE Access · doi:10.1109/access.2022.3160170 · Cited by 10
Anomaly Detection for CPS via Memory-Augmented Reconstruction and Time Series Prediction
Sun, Z., & Li, J. (2022) · IEEE International Conference on Mobile Ad-Hoc and Sensor Systems (MASS) · doi:10.1109/mass56207.2022.00080 · Cited by 8
“OmniAbnormal and USAD capture temporal information with sequential inputs and use probabilistic generative methods or adversarial training, respectively, to make the reconstruction more efficient and robust.”
BackgroundMethodInfluentialMadSGM: Multivariate Anomaly Detection with Score-based Generative Models
Lim, H., Park, S., Kim, M., Lee, J., Lim, S., & Park, N. (2023) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3583780.3614956 · Cited by 12
Anomalous Signal Detection for Cyber-Physical Systems Using Interpretable Causal Neural Network
Zhang, S., & Liu, J. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10096634 · Cited by 7
Tan, G., Chen, P., & Li, M. (2023) · IEEE International Conference on Software Quality, Reliability and Security (QRS) · doi:10.1109/qrs60937.2023.00011 · Cited by 8
“Anomaly detection is also a classification problem that involves distinguishing between normals and anomalies [22].”
InfluentialZhang, Z., Zhao, L., Gao, M., Chen, Y., Wang, J., & Wang, C. (2024) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2024.108287 · Cited by 8
Separable contextual graph neural networks to identify tailgating-oriented traffic congestion
Lee, J., & Lee, S. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.124354 · Cited by 8
Yu, B., Yu, Y., Xiang, G., & Lin, R. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3460930 · Cited by 7
“These included traditional method-based models: Isolation Forest [5] and Deep-SVDD [22]; deep-learning-based models: LSTM-VAE [23], OmniAnomaly [25], and USAD [26]; graph neural-network-based models: GDN [13] and MTAD-GAT [29]; and the latest SOTA models: AMSL [27] and AnomalyTransformer [28].”
BackgroundMethodStatioCL : Contrastive Learning for Time Series via Non-Stationary and Temporal Contrast
Wu, Y. Y., Dang, T., Spathis, D., Jia, H., & Mascolo, C. (2024) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3627673.3679732 · Cited by 11
Han, C. S., Kim, H., & Lee, K. M. (2025) · Sensors · doi:10.3390/s25082510 · Cited by 9
Chen, L., Tang, J., Zou, Y. Y., Liu, X., Xie, X., & Deng, G. (2025) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3565807 · Cited by 10
“In addition, the USAD model [20] uses two multilayer MLPs to design an autoencoder network for anomaly detection.”
Tian, J., Li, M., Fang, L., & Chen, L. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3588674 · Cited by 9
“[26], USAD (2020) [7], NSIBF (2021) [8], FuSAGNet (2022) [9], FuGLAD (2024) [27], the supervised RobustTAD (2020) [11] and CutAddPaste (2024) [12], and the self-supervised CTAD (2023) [14] and DCdetector (2023) [15].”
InfluentialAnomaly Detection for UAV Flight Data via Reconstruction–Prediction Co-Learning Attention Network
Zhou, Z., Zhong, J., Zhang, Y., & Miao, Q. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3593530 · Cited by 12
“…in this study is compared with prediction-based algorithms: LSTM [31], TSMixer [32], iTransformer [33], TimesNet [34], reconstruction-based algorithms: LSTM-VAE [35], deep autoencoding Gaussian mixture model (DAGMM) [36], and dual-task algorithms: MTAD-GAT [37], USAD-M [38], and MSCRED [39].”
Dual Graph Learning for Multivariate Time Series Anomaly Detection in IoUT
Luo, Y., Chen, J., Qin, S., & Wu, D. O. (2025) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2025.3598425 · Cited by 8
“This includes both graph-based techniques, such as MTAD-GAT [26], GDN [27], GRELEN [50] GSC-MAD [59], PS-GDNML [60] and GTA [24], as well as non-graph-based approaches like LSTM-VAE [18], MAD-GAN [19], Omni-Anomaly [20], USAD [61], CAE-AD [21], DiG [62] and TranAD [23].”
xLSTMAD: A Powerful xLSTM-based Method for Anomaly Detection
Faber, K., Pietroń, M., Żurek, D., & Corizzo, R. (2025) · Proceedings · doi:10.1109/icdm65498.2025.00032 · Cited by 20
Anomaly Detection in Industrial Multivariate Time-Series Data With Neutrosophic Theory
Liu, P., Han, Q., Wu, T., & Tao, W. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3262612 · Cited by 8
“Audibert et al. [18] proposed an encoder-decoder architecture within an adversarial training framework, this method added a decoder based on AE and conducted adversarial training through two AEs.”
MethodconDENSE: Conditional Density Estimation for Time Series Anomaly Detection
Moore, A. K., & Morelli, D. (2024) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.14849 · Cited by 8
“…extreme value analysis is typically used to define a threshold on this anomaly score (Deng & Hooi, 2021; Benecki, et al., 2021; Su, et al., 2019; Tuli et al., 2022; Audibert, et al., 2020; Boniol, et al., 2021; Kingsbury & Alvaro, 2020; Li, et al., 2019; Zhang, et al., 2019; Zhao, et al., 2020).”
MethodInterpretability of Causal Discovery in Tracking Deterioration in a Highly Dynamic Process
Choudhary, A., Vuković, M., Mutlu, B., Haslgrübler, M., & Kern, R. (2024) · Sensors · doi:10.3390/s24123728 · Cited by 10
“The performance of USAD in detecting anomalies in multivariate time series data has been evaluated on five publicly available datasets [32].”
BackgroundMethodInfluentialAD-NEv: A Scalable Multilevel Neuroevolution Framework for Multivariate Anomaly Detection
Pietroń, M., Żurek, D., Faber, K., & Corizzo, R. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3439404 · Cited by 14
“Among recent works on semi-supervised and unsupervised multivariate time-series anomaly detection, deep learning-based methods achieve the best results on well-known benchmarks [2], [3], [4], [5], [6], [7], [8], [9].”
Anomaly detection in sensor data via encoding time series into images
Ma, J., & Wang, H. (2024) · Journal of King Saud University - Computer and Information Sciences · doi:10.1016/j.jksuci.2024.102232 · Cited by 9
Chen, X., Kou, J., Li, H., Zhang, Y., Ma, J., Li, C., & Tu, B. (2025) · Computers & Security · doi:10.1016/j.cose.2025.104461 · Cited by 8
Unsupervised Anomaly Detection Approach for Multivariate Time Series
Zhou, Y., Song, Y., & Qian, M. (2021) · IEEE International Conference on Software Quality, Reliability and Security Companion (QRS-C) · doi:10.1109/qrs-c55045.2021.00042 · Cited by 9
Zong, W., Zhou, F., Pavlovski, M., & Qian, W. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-00126-0_37 · Cited by 8
Towards Real-Time Safety Analysis of Small Unmanned Aerial Systems in the National Airspace
Cleland‐Huang, J., Chawla, N. V., Cohen, M. B., Islam, M. N. A., Sinha, U., Spirkovska, L., Ma, Y., Purandare, S., & Chowdhury, M. T. (2022) · AIAA AVIATION Forum · doi:10.2514/6.2022-3540 · Cited by 7
Sha, Y., Gou, S., Faber, J., Liu, B., Li, W., Schramm, S., Stöcker, H., Steckenreiter, T., Vnucec, D., Wetzstein, N., et al. (2022) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3534678.3539133 · Cited by 19
“Following previous researches [23, 24, 25], we select the AUC score of ROC curve area [26] as an evaluation metric to determine the degree of separability of samples.”
MethodHu, Z., Chen, P., Yu, G., He, Z., & Li, X. (2022) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws55610.2022.00031 · Cited by 9
“Compared with state-of-the-art methods [4], [5], [7], TS-InvarNet can perform with less storage as Table III shows.”
BackgroundMethodInfluentialGe, N., Weng, X., & Yang, Q. (2022) · Applied Intelligence · doi:10.1007/s10489-022-04337-y · Cited by 8
“As a result, it performs worse than USAD on WADI and MSL. GDN, MTAD-GAT, and GTA are methods based on GNNs.”
BackgroundMethodInfluentialAnomaly Detection for Process Industry using Class-Dependent Temperatures Loss
Li, Z., Yang, Z., Tang, D., Peng, S., & Li, J. (2023) · International Conference on Computer and Communication Systems (ICCCS) · doi:10.1109/icccs57501.2023.10150495 · Cited by 7
A Stitch in Time Saves Nine: Enabling Early Anomaly Detection with Correlation Analysis
Ang, Y., Huang, Q., Tung, A. K. H., & Huang, Z. (2023) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde55515.2023.00143 · Cited by 17
“…three efficient data mining-based methods LOF [15], ECOD [48], and IForest [50], two state-of-the-art unsupervised deep learning-based methods USAD [9] and RCoders [2], [3], and four advanced univariate anomaly detection methods Series2Graph (S2G) [13], SAND [14], its online version SAND ∗ , and…”
BackgroundMethodInfluentialCauseFormer: Interpretable Anomaly Detection With Stepwise Attention for Cloud Service
Zhong, G., Liu, F., Jiang, J., & Chen, C. L. P. (2023) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2023.3299846 · Cited by 9
“Inversely, the USAD [16] updates the network framework by adding one more decoder as the discriminator.”
MethodTime Series Anomaly Detection with Reconstruction-Based State-Space Models
Wang, F., Wang, K., & Yao, B. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-44213-1_7 · Cited by 8
“• UnSupervised Anomaly Detection (USAD) [16] adversarially trains an autoencoder model to amplify the reconstruction errors of abnormal samples”
BackgroundInfluentialA Federated Learning Approach for Efficient Anomaly Detection in Electric Power Steering Systems
Kea, K., Han, Y., & Min, Y.-J. (2024) · IEEE Access · doi:10.1109/access.2024.3397000 · Cited by 12
ALAE: self-attention reconstruction network for multivariate time series anomaly identification
Jiang, K., Liu, H., Ruan, H., Zhao, J., & Lin, Y. (2023) · Soft Computing · doi:10.1007/s00500-023-08467-4 · Cited by 6
“Alternatively, non-linear dimensionality reduction techniques can be used to extract high-dimensional data features for data reconstruction (Amarbayasgalan et al. 2020; Audibert et al. 2020).”
MethodGenerative Optimization Networks for Memory Efficient Data GenerationPreprint
Tuli, S., Tuli, S., Casale, G., & Jennings, N. R. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2110.02912 · Cited by 11
“Other methods use generative models like variational autoencoders or GANs for the same task and have been shown to perform very well [1, 13].”
BackgroundMethodInfluentialXiao, Q., Shao, S., & Wang, J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2110.08306 · Cited by 12
“We choose convolutional networks rather than fully connected networks, e.g. (Xu et al. 2018; Su et al. 2019; Audibert et al. 2020), because they are more computationally effective for large datasets.”
BackgroundMethodInfluentialYan, P., Abdulkadir, A., Luley, P.-P., Rosenthal, M., Schatte, G. A., Grewe, B. F., & Stadelmann, T. (2023) · Zürcher Hochschule für Angewandte Wissenschaften digital collection (Zurich University of Applied Sciences) · doi:10.48550/arxiv.2307.05638 · Cited by 11
“Audibert et al. propose a fast and stable method – Unsupervised anomaly detection for multivariate time series [92], based on adversely trained autoencoders.”
MethodAnomaly Detection based on Robust Spatial-temporal Modeling for Industrial Control Systems
Li, S., Liu, J., Pan, Z., Lv, S., Si, S., & Sun, L. (2022) · 2022 IEEE 19th International Conference on Mobile Ad Hoc and Smart Systems (MASS) · doi:10.1109/mass56207.2022.00058 · Cited by 6
“Figure 6(c) summarizes the obtained results using 5 different head numbers h ∈ [2, 4, 8, 16, 32].”
BackgroundMethodInfluentialSGFM: Conditional Flow Matching for Time Series Anomaly Detection With State Space Models
He, Y., Yan, T., Zhan, Y., Feng, Z., & Xia, Y. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3439672 · Cited by 6
“1) USAD: USAD [14] shows great stability in time series anomaly detection by combining autoencoders and adversarial training.”
BackgroundMethodInfluentialIdentifying Performance Issues in Cloud Service Systems Based on Relational-Temporal Features
Gu, W., Liu, J., Chen, Z., Zhang, J., Su, Y., Gu, J., Feng, C. Q., Yang, Z., Yang, Y., & Lyu, M. R.-T. (2025) · ACM Transactions on Software Engineering and Methodology · doi:10.1145/3702978 · Cited by 16
“F1 score is the harmonic mean of the precision and recall, which symmetrically represents both precision and recall in one metric.”
Detecting Multivariate Time Series Anomalies With Cascade Decomposition Consistency
Li, R., Liu, Z., Zhu, X., Li, L., & Cao, X. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3547479 · Cited by 9
“4) USAD [55]: A reconstruction-based method employs a two-phase autoencoder architecture, detecting anomalies based on reconstruction deviations.”
Multimodal Time-Frequency Pseudo Anomalies for Atrial Fibrillation Anomaly Detection
Fan, H., Han, H., Chang, H., & Wang, Z. (2025) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2025.3547531 · Cited by 6
CiTranGAN: Channel-Independent Based-Anomaly Detection for Multivariate Time Series Data
Chen, X., Li, T., Ma, Z., Chen, J., Guo, J., & Liu, Z. (2025) · Electronics · doi:10.3390/electronics14091857 · Cited by 6
Quantized Auto Encoder-Based Anomaly Detection for Multivariate Time Series Data in 5G Networks
Trappolini, G., Purificato, A., Siciliano, F., D’Addona, L., Spagnolo, A. M., Dato, D., & Silvestri, F. (2025) · IEEE Access · doi:10.1109/access.2025.3568133 · Cited by 8
ADAMAS: Adaptive Domain-Aware Performance Anomaly Detection in Cloud Service Systems
Gu, W., Gu, J., Liu, J., Chen, Z., Zhang, J., Kuang, J., Feng, C., Yang, Y., & Lyu, M. R.-T. (2025) · ACM/IEEE International Conference on Software Engineering: Companion Proceedings (ICSE-Companion) · doi:10.1109/icse55347.2025.00084 · Cited by 10
“Thus, we adopt the widely used evaluation strategy following [11], [46], [57], [58], [24].”
Inter-layer explainable variational autoencoder model for multivariate time series anomaly detection
Zhang, X., Wang, G., Chen, Y., Yang, W., & Wang, G. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.111585 · Cited by 9
Machine Learning Supported Detection of Incoupling Interfering Signals Through Autoencoders
Cahani, I., Ueltzen, R., Elsayed, M. O. K., Kampert, E., & Stiemer, M. (2025) · International Symposium on Electromagnetic Compatibility - EMC Europe (EMC Europe) · doi:10.1109/emceurope61644.2025.11176344 · Cited by 6
Al-Omari, T. (2025) · Engineering Technology & Applied Science Research · doi:10.48084/etasr.13590 · Cited by 7
Galphade, M., Nikam, V. B., Banerjee, B., More, N., Kiwelekar, A. W., & Sharma, P. (2026) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2026.114700 · Cited by 6
TCAD: Unsupervised Anomaly Detection Based on Global Local Representation Differences
Duan, Y., Xiang, M., Bingtao, Z., Fu, D., & Liu, H. (2022) · IEEE Access · doi:10.1109/access.2022.3216930 · Cited by 7
“Benefiting from the improved learning the ability of neural network temporal association representation, some recent deep learning models, such as USAD [7], OmniAnomaly [8], MAD-GAN [9], LSTM-AE [10], VAR [11], ConvLSTM [12], LSTM-VAE [13], have achieved superior results.”
BackgroundInfluentialAdaptive Thresholding Heuristic for KPI Anomaly Detection
Isaac, E. R. H. P., & Sharma, A. (2024) · International Conference on Communication Systems & Networks (COMSNETS) · doi:10.1109/comsnets59351.2024.10427016 · Cited by 10
“AD solutions based on Deep Learning (DL) [4], [5], [6], [7] are computationally expensive.”
MethodYi, S., Zheng, S., Yang, S., Zhou, G., & He, J. (2024) · Annals of Nuclear Energy · doi:10.1016/j.anucene.2024.110557 · Cited by 6
Towards Efficient Deep Autoencoders for Multivariate Time Series Anomaly Detection
Pietroń, M., Żurek, D., Faber, K., & Corizzo, R. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-63759-9_46 · Cited by 7
Lee, Y., Song, S., Park, K.-Y., Koo, B.-M., & Baek, J. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3507042 · Cited by 6
Cao, H., Xu, G., He, Z., Shi, S., Xu, S., Wu, C., & Ning, J. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3563069 · Cited by 6
Anomaly Detection for Hydraulic Systems under Test
Neufeld, D., & Schmid, U. (2021) · IEEE International Conference on Emerging Technologies and Factory Automation (ETFA) · doi:10.1109/etfa45728.2021.9613265 · Cited by 6
Unsupervised Anomaly Detection by Autoencoder with Feature Decomposition
Guo, Y., Zhu, X., Hu, Z., & Zhan, Z. (2022) · International Conference on Machine Learning and Computing (ICMLC) · doi:10.1145/3529836.3529924 · Cited by 6
“For example, search engines sometimes need to identify and filter abnormal pictures without prior information when retrieving pictures [29]; In some industrial scenarios, one needs to monitor and maintain sensors without expert experience and clear abnormal definition [1].”
BackgroundAnomaly Detection Method for Time Series Data Based on Transformer Reconstruction
Wang, Y., & Li, J. (2023) · International Conference on Informatics, Environment, Energy and Applications (IEEA) · doi:10.1145/3594692.3594702 · Cited by 6
“The Transformer model, on the other hand, allows a single inference of the complete input sequence using positional encoding, and the time of training and inference is almost independent of the sequence length [5].”
MethodHajisafi, A., Lin, H., Shaham, S., Hu, H., Siampou, M. D., Chiang, Y., & Shahabi, C. (2023) · ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL GIS) · doi:10.1145/3589132.3625567 · Cited by 6
“In contrast, multivariate time series techniques aim to capture interactions and co-movements among a group of variables [1, 34].”
BackgroundContext-Aware Deep Time-Series Decomposition for Anomaly Detection in Businesses
Nam, Y., Trirat, P., Kim, T.-Y., Lee, Y., & Lee, J.-G. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43427-3_20 · Cited by 10
Xue, B., Gao, X., Zhai, F., Li, B., Yu, J., Fu, S., Chen, L., & Meng, Z. (2023) · Applied Intelligence · doi:10.1007/s10489-023-04985-8 · Cited by 7
“…adjustment strategy (PA) and the point-adjusted F1-score (F1 PA ) are also proposed and applied for anomalous events detection [29, 43, 44], but the F1 PA does not represent the real performance of the model, which might give a high score even when a large number of events are not detected [26].”
BackgroundMethodGCFormer: Granger Causality based Attention Mechanism for Multivariate Time Series Anomaly Detection
Xing, S., Niu, J., & Ren, T. (2023) · Proceedings · doi:10.1109/icdm58522.2023.00187 · Cited by 7
“Prediction-based models [2], [7] detect anomalies by evaluating prediction errors, while reconstruction-based models [3], [4] rely on examination of reconstruction errors.”
BackgroundMethodResultXi, Z., Zhou, J., Yang, B., Zhang, Y., Zhang, Z., & Li, D. (2024) · Measurement · doi:10.1016/j.measurement.2024.114482 · Cited by 6
A Robust Prioritized Anomaly Detection when Not All Anomalies are of Primary Interest
Lu, G., Zhou, F., Pavlovski, M., Zhou, C., & Jin, C. (2024) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde60146.2024.00065 · Cited by 10
Sun, Y., Liang, M., Zhang, S., Che, Z., Luo, Z., Li, D., Zhang, Y., Pei, D., Pan, L., & Hou, L. (2025) · ACM Transactions on Software Engineering and Methodology · doi:10.1145/3702984 · Cited by 8
“How much do the key techniques contribute to the overall performance?”
Li, J. W., Yu, Z., Jiang, Q., & Cao, Z. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3548251 · Cited by 5
“Anomaly detection in MTS refers to the identification of data that deviate from the expected distribution within a collection of interconnected time series [5].”
Ye, F., Zhang, K., Sun, J., & Li, N. (2025) · International Journal of Intelligent Systems · doi:10.1155/int/4370827 · Cited by 5
Yu, Y.-C., Ouyang, Y., & Lin, C.-A. (2025) · IEEE Access · doi:10.1109/access.2025.3610684 · Cited by 9
A time–frequency contrastive learning model for anomaly detection in multivariate time series
Zhāng, W., Li, X., Li, J., Ma, J., Pengfei, K., Zhang, S., & Liu, Y. (2025) · Complex & Intelligent Systems · doi:10.1007/s40747-025-02119-w · Cited by 8
Memory-Efficient Artificial Intelligence Framework for Real-Time Multivariate Anomaly Detection
Hwang, J., Kim, S., & Park, S. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3653288 · Cited by 5
“…neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory (LSTM) networks, transformers, and graph neural networks (GNNs), for modeling complex, nonlinear, and temporal multivariate dependencies in sensor data [17], [24], [25], [26], [27], [28], [29], [30], [31], [32].”
Kong, H., Li, T., Ge, J., Zhang, L., & Li, L. (2024) · Automated Software Engineering · doi:10.1007/s10515-024-00445-w · Cited by 9
“Audibert et al. (2020) introduced a fast and stable unsupervised anomaly detection method using adversarial training of autoencoders.”
Kim, B.-S., & Ro, S.-K. (2024) · International Journal of Precision Engineering and Manufacturing · doi:10.1007/s12541-024-01039-y · Cited by 5
Multivariate Time-Series Anomaly Detection in IoT with a Bi-Dual GM GRU Autoencoder
Yu, Y.-C., Ouyang, Y., Wu, L.-W., Lin, C.-A., & Tsai, K. (2024) · IEEE Annual International Computer Software and Applications Conference (COMPSAC) · doi:10.1109/compsac61105.2024.00106 · Cited by 6
“Inspired by the adversarial concept in GANs, Audibert et al. [21] applied similar adversarial training techniques based on AE framework by employing a two-phase training process, which enables two AE models to learn both the reconstruction of input time series and the discrimination between real…”
MethodSTformer: Spatio-Temporal Transformer for Multivariate Time Series Anomaly Detection
Li, Z., Zhang, H., & Zheng, W. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72347-6_20 · Cited by 5
Graph Laplacian Eigenvalues Empowered VAEs: A Novel Approach to Adaptive Latent Dimension Choice
Saravana, M. K., Roopa, M. S., Arunalatha, J. S., & Venugopal, K. R. (2024) · IEEE Access · doi:10.1109/access.2024.3460971 · Cited by 8
Anomaly sign detection by monitoring thousands of process values using a two-stage autoencoder
Naito, S., Taguchi, Y., Kato, Y., Nakata, K., Miyake, R., Nagura, I., Tominaga, S., & Aoki, T. (2021) · Mechanical Engineering Journal · doi:10.1299/mej.20-00534 · Cited by 5
Unsupervised Tensor based Feature Extraction and Outlier Detection for Multivariate Time Series
Matsue, K., & Sugiyama, M. (2021) · IEEE International Conference on Data Science and Advanced Analytics (DSAA) · doi:10.1109/dsaa53316.2021.9564117 · Cited by 6
Forecast-based Multi-aspect Framework for Multivariate Time-series Anomaly Detection
Wang, L., Lin, Y., Wu, Y., Chen, H., Wang, F., & Yang, H. (2021) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata52589.2021.9671776 · Cited by 8
“Among the vast literature on time series anomaly detection over the past decade [19], [20], [21], [7], [22], [23], [6], one can divide them into three groups: (1) proximity-based; (2) reconstruction-based; and (3) forecast-based.”
BackgroundLSTM-AE for Anomaly Detection on Multivariate Telemetry Data
Abdennebi, A., Tunçay, A., Yılmaz, C., Koyuncu, A., & Güngör, O. H. (2023) · IEEE/ACIS International Conference on Software Engineering Research, Management and Applications (SERA) · doi:10.1109/sera57763.2023.10197673 · Cited by 8
“We noticed that some literature models [1] [5] are pruned only to detect anomalies in offline data which differs from our primary objective.”
BackgroundSun, Y., Liang, M., Che, Z., Li, D., Zheng, T., Zhang, S., Zhu, P., Zhang, Y., & Pei, D. (2023) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws60048.2023.00033 · Cited by 5
Toward Efficient Traffic Incident Detection via Explicit Edge-Level Incident Modeling
Liu, C., Chen, J., Liu, H., Li, S., & He, S. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3371482 · Cited by 6
“Previous studies have commonly employed a single anomaly score, defined based on either the reconstruction error or prediction error of sensor measurements, for the detection of both types of incidents [25], [26], [27].”
MethodHu, M. C., Zhang, F., & Wu, H. (2024) · Applied Sciences · doi:10.3390/app14052202 · Cited by 5
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.