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 201–300 of 1,000
Entropy Causal Graphs for Multivariate Time Series Anomaly Detection
Febrinanto, F. G., Moore, K., Thapa, C., Liu, M., Saikrishna, V., Ma, J., & Xia, F. (2025) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3757922 · Cited by 26
Active-MTSAD: Multivariate Time Series Anomaly Detection With Active Learning
Wang, W., Chen, P., Xu, Y., & He, Z. (2022) · International Conference on Dependable Systems & Networks/Proceedings · doi:10.1109/dsn53405.2022.00036 · Cited by 23
Uncertainty-informed dynamic threshold for time series anomaly detection
Lee, J., Lee, J., & Kim, S. B. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127379 · Cited by 23
Dou, H., Shi, P., Zhang, Y., Chen, P., & Zheng, Z. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3570337 · Cited by 17
“…can be further categorized into the prediction-based methods: MTAD-GAT [8], and GDN [9]; the reconstruction-based methods: BeatGAN [42], USAD [43], OmniAnomaly [44], InterFusion [11], TranAD [12], ATF-UAD [16], DTAAD [45], and CATCH [46]; the decomposition-based method DecompTran [14]; the…”
Two-Stage Deep Anomaly Detection With Heterogeneous Time Series Data
Jeong, K.-J., Park, J.-D., Hwang, K., Kim, S., & Shin, W.-Y. (2022) · IEEE Access · doi:10.1109/access.2022.3147188 · Cited by 19
“To demonstrate the validity of our T-DAD framework on other datasets, we present the performance comparison between T-DAD and three single-stage benchmark approaches (including DAE, DAGMM, and USAD) in Table 4, where the DAE and LSTM-DAE models are used in Stages I and II, respectively, of our T-DAD framework.”
BackgroundMethodInfluentialLundström, A., O’Nils, M., Qureshi, F. Z., & Jantsch, A. (2022) · IEEE Access · doi:10.1109/access.2022.3213038 · Cited by 16
“These are point-wise score and point adjust score [8].”
BackgroundMethodInfluentialBui, N.-T., Hoang, D.-H., Phan, T., Tran, M., Patel, B. D., Adjeroh, D., & Le, N. (2024) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi56570.2024.10635676 · Cited by 18
“We [6] 0.775 USAD (SIGKDD’20) [7] 0.785 TranAD (VLDB’22) [12] 0.788 Anomaly Transformer (ICLR’22) [8] 0.762 Zheng et al. (MICCAI’22) [18] 0.757 BeatGAN (TKDE’22) [13] 0.799 Jiang et al. (MICCAI’23) [14] 0.860 TSRNet (Ours) 0.860 compare the result when inferring with Peak-based Error to the result…”
MethodResultInfluentialYao, M., Tao, D., Qi, P., & Gao, R. (2024) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2024.3486688 · Cited by 18
“Some prior works point out that the anomaly portion may vary in actual practice [15], [16], we therefore conduct a case study to investigate the impacts of anomaly portion in the overall AD performance.”
InfluentialFlight anomaly detection and localization based on flight data fusion and random channel masking
Zhong, J., Zhang, H., & Miao, Q. (2025) · Applied Soft Computing · doi:10.1016/j.asoc.2025.113023 · Cited by 17
Efficient Model Monitoring for Quality Control in Cardiac Image Segmentation
Galati, F., & Zuluaga, M. A. (2021) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-78710-3_11 · Cited by 15
“‖X̂ ′ − X̂‖2, to quantify how anomalous is a sample [1].”
Self-citationWang, X., Yin, K., Ouyang, Q., Wen, X., Zhang, S., Zhang, W., Cao, L., Han, J., Jin, X., & Pei, D. (2022) · Proceedings/Proceedings - International Symposium on Software Reliability Engineering · doi:10.1109/issre55969.2022.00043 · Cited by 20
Zhao, M., Peng, H., & Li, L. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3520362 · Cited by 22
“Anomaly detection methods based on AEs [7], [8], [9], [10], [11], [12] use various networks (such as LSTM, RNNs, and CNN) to create encoders and decoders.”
InfluentialTSB-AutoAD: Towards Automated Solutions for Time-Series Anomaly Detection
Liu, Q., Lee, S., & Paparrizos, J. (2025) · Proceedings of the VLDB Endowment · doi:10.14778/3749646.3749699 · Cited by 28
Diffusion models with self-conditioning guidance for multivariate time series anomaly detection
Li, Y., Wen, Z., Chen, Z., Mei, J., Lin, M., & Zhu, M. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.114511 · Cited by 22
TFDPM: Attack detection for cyber–physical systems with diffusion probabilistic models
Yan, T., Zhou, T., Zhan, Y., & Xia, Y. (2022) · Knowledge-Based Systems · doi:10.1016/j.knosys.2022.109743 · Cited by 15
“USAD : USAD [47] is a recently proposed method based on autoencoders.”
Decomposed Transformer with Frequency Attention for Multivariate Time Series Anomaly Detection
Qin, S., Zhu, J., Wang, D., Ou, L. S., Gui, H., & Tao, G. (2022) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata55660.2022.10021063 · Cited by 22
“Method SMD SMAP MSL SWaT PSMP R F1 P R F1 P R F1 P R F1 P R F1 LSTM-VAE 77.63 89.12 82.98 88.65 72.18 79.57 92.32 78.61 84.92 81.78 78.55 80.13 90.16 74.48 81.57 OmniAnomaly 83.34 94.49 88.57 74.16 97.76 84.34 88.67 91.17 89.89 86.33 76.94 81.36 91.61 71.36 80.23 MTAD-GAT 88.28 84.92 86.57 89.06 91.23 90.13 87.54 94.4…”
MethodInfluentialAutoKAD: Empowering KPI Anomaly Detection with Label-Free Deployment
Yu, Z.-Y., Pei, C.-H., Zhang, S.-L., Wen, X., Li, J.-H., Xie, G.-G., & Pei, D. (2023) · IEEE International Symposium on Software Reliability Engineering · doi:10.1109/issre59848.2023.00063 · Cited by 15
Asymptotic Consistent Graph Structure Learning for Multivariate Time-Series Anomaly Detection
Pang, H., Wei, S., Li, Y., Liu, T., Zhang, H., Qin, Y. H., & Zhao, Y. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3369159 · Cited by 17
“These baseline methods are as follows: LSTM-NDT [10], DAGMM [37], OmniAnomaly [18], MSCRED [15], MAD-GAN [38], USAD [36], MTAD-GAT [27], CAE-M [39], InterFusion [40], GDN [1], TranAD [7], CrossFuN [20], DyGraphAD [16], and Anomaly Transformer [23].”
MethodGiannoulidis, A., Gounaris, A., Naskos, A., Nikolaidis, N., & Caljouw, D. (2024) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-024-02352-z · Cited by 22
“Using the reconstruction error of data for a PdM application, when sufficient amount of normal data are collected, can be done using state-of-the art methods like Audibert et al. (2020), Xu et al. (2018), Tuli et al. (2022) too.”
Alrumaih, T. N. I., & Alenazi, M. J. F. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3390691 · Cited by 19
“Similarto[16], an unsupervised multivariate anomaly detection using GAN (MAD-GAN) model was proposed in [18].”
BackgroundMethodLe, T. M. V., Vu, H. C., Ponchet-Durupt, A., Boudaoud, N., Cherfi-Boulanger, Z., & Trang, T. N. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.127791 · Cited by 17
Self-supervised Spatial-Temporal Normality Learning for Time Series Anomaly Detection
Chen, Y., Xu, H., Pang, G., Qiao, H., Zhou, Y., & Shang, M. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-70365-2_9 · Cited by 21
AssureSense: A Framework for Enabling Sensor Fault Detection in Low-Power IoT Edge Devices
Attarha, S., & Förster, A. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3451349 · Cited by 19
“Additionally, we implemented the USAD [33] model.”
BackgroundYao, M., Tao, D., Gao, R., & Qi, P. (2024) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2024.3421600 · Cited by 17
“Existing methods [9], [10], [11], [12], [13], [15], [16] generally rely on reconstruction errors to identify anomalies.”
Toward Intelligent Attack Detection With Causal Transformer in Internet of Things
Zeng, Z., Zhao, B., Deng, X., Liu, X., Zheng, J., & Chen, J. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3512531 · Cited by 16
“…apply. accuracy, precision and F1 with previous benchmark methods, such as TranAD [22], LSTM_Univariate [43], Attention [26], LSTM_AD [27], DAGMM [44], OmniAnomaly [38], USAD [45], MSCRED [46], MTAD_GAT [47], GDN [48], MAD_GAN [49], isolation forest [50], SVM [17], LOF [51], and LSTM [25].”
A Survey of Change Point Detection in Dynamic Graphs
Zhou, Y., Gao, S. B., Guo, D., Wei, X., Rokne, J. G., & Wang, H. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2024.3523857 · Cited by 25
EST transformer: enhanced spatiotemporal representation learning for time series anomaly detection
Gao, Y., Su, R., Ben, X., & Chen, L. (2025) · Journal of Intelligent Information Systems · doi:10.1007/s10844-025-00918-8 · Cited by 14
“Identifying anomalies in multivariate time series data is quite challenging because it requires consideration of both the temporal context dependencies and the interdependencies among different dimensions (Audibert et al., 2020).”
Incremental Semi-Supervised Learning for Data Streams Classification in Internet of Things
Jiang, J., Wang, B., Tang, Q., Zhong, G., Tang, X., & Rodrigues, J. J. P. C. (2025) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2025.3546649 · Cited by 14
“USAD [57]: USAD is a multivariate time series method for anomaly detection based on adversarial trained autoencoders.”
Kim, S., Lee, T.-H., & Lee, J. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127001 · Cited by 18
Zhang, J., Wang, X., Yang, Y., Miao, H., & Yang, S. (2025) · Journal of King Saud University - Computer and Information Sciences · doi:10.1007/s44443-025-00024-3 · Cited by 20
“In previous work, the point-adjusted (PA) (Audibert et al. 2020) method was used to evaluate multivariate time series anomaly detection models.”
From Univariate to Multivariate Time Series Anomaly Detection with Non-Local Information
Audibert, J., Marti, S., Guyard, F., & Zuluaga, M. A. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-91445-5_12 · Cited by 14
“We implement the AE using Pytorch and we used publicly available implementations for MP[1]4, LSTM-VAE2 and USAD3.”
BackgroundMethodInfluentialSelf-citationMulti-scale Cross-restoration Framework for Electrocardiogram Anomaly Detection
Jiang, A., Huang, C., Cao, Q., Wu, S., Zeng, Z., Chen, K., Zhang, Y., & Wang, Y. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43907-0_9 · Cited by 17
Detecting Cloud Anomaly via Broad Network-Based Contrastive Autoencoder
Zhong, G., Liu, F., Jiang, J., Wang, B., Yao, X., & Chen, C. L. P. (2024) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2024.3353772 · Cited by 15
“However, for more comprehensive evaluation [21], [53], we import the macro R (¯ R ) , macro P (¯ P ) , and macro F1 ( ¯ F 1) in this paper.”
BackgroundMethodInfluentialA Deep Deterministic Policy Gradient-Based Method for Enforcing Service Fault-Tolerance in MEC
Long, T., Chen, P., Xia, Y., Ma, Y., Sun, X., Zhao, J., & Lyu, Y. (2024) · Chinese Journal of Electronics · doi:10.23919/cje.2023.00.105 · Cited by 18
Ghanim, J., & Awad, M. (2024) · Journal of Artificial Intelligence and Soft Computing Research · doi:10.2478/jaiscr-2025-0001 · Cited by 14
Hojjati, H., Sadeghi, M., & Armanfard, N. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43430-3_15 · Cited by 29
Wang, Z., Liu, H., Kuang, L., Zhang, X., Chen, X., & Du, J. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.112775 · Cited by 14
He, Z., Chen, P., & Huang, T. (2022) · Proceedings/Proceedings - International Symposium on Software Reliability Engineering · doi:10.1109/issre55969.2022.00014 · Cited by 16
“Therefore, it has been widely adopted by many previous deep unsupervised models [12], [14], [16], [27], [28].”
BackgroundMethodInfluentialMENDEL: Time series anomaly detection using transfer learning for industrial control systems
Park, J., Kim, B., & Kim, H. (2023) · IEEE International Conference on Big Data and Smart Computing (BigComp) · doi:10.1109/bigcomp57234.2023.00049 · Cited by 12
RESIST: Robust Transformer for Unsupervised Time Series Anomaly Detection
Najari, N., Berlemont, S., Lefebvre, G., Duffner, S., & García, C. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-24378-3_5 · Cited by 16
Masked Graph Neural Networks for Unsupervised Anomaly Detection in Multivariate Time Series
Xu, K., Li, Y., Li, Y., Xu, L., Li, R., & Dong, Z. (2023) · Sensors · doi:10.3390/s23177552 · Cited by 16
“USAD: USAD [43] is an unsupervised anomaly detection approach for multivariate time series based on a GAN-inspired self-encoder architecture.”
MethodInfluentialStehle, F. K., Vandelli, W., Zahn, F., Avolio, G., & Fröning, H. (2024) · International Conference on Supercomputing · doi:10.1145/3650200.3656637 · Cited by 20
“These three non-transformer candidates are OmniAnomaly [54], DAGMM [63], and USAD [10].”
BackgroundMethodLi, Q., Li, K., Gao, X., Fu, J., & Zhang, L. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3436113 · Cited by 22
Wang, L., Wang, X., Zhang, J., Wang, J., & Yu, H. (2024) · Structures · doi:10.1016/j.istruc.2024.107148 · Cited by 12
Naito, S., Taguchi, Y., Nakata, K., & Kato, Y. (2021) · IEEE ... International Conference on Data Mining workshops · doi:10.1109/icdmw53433.2021.00072 · Cited by 13
“In [7], we suppose that this is one of the main reasons for the small F1 score in the WADI dataset.”
BackgroundMethodInfluentialXie, X., Ning, W., Huang, Y., Li, Z., Yu, S., & Yang, H. (2022) · International Journal of Intelligent Systems · doi:10.1002/int.23027 · Cited by 13
Agent based dynamic thresholding for adaptive anomaly detection using reinforcement learning
Yang, X., Howley, E., & Schukat, M. (2025) · Neural Computing and Applications · doi:10.1007/s00521-024-10536-0 · Cited by 11
“In the literature, many existing anomaly detection approaches use static or expert-defined thresholds [3, 5, 6].”
InfluentialAn Effective WGAN-Based Anomaly Detection Model for IoT Multivariate Time Series
Qi, S., Chen, J., Chen, P., Wen, P., Shan, W., & Xiong, L. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-33374-3_7 · Cited by 12
Practical Anomaly Detection over Multivariate Monitoring Metrics for Online Services
Liu, J., Yang, T., Chen, Z., Su, Y., Feng, C., Yang, Z., & Lyu, M. R.-T. (2023) · Proceedings/Proceedings - International Symposium on Software Reliability Engineering · doi:10.1109/issre59848.2023.00045 · Cited by 16
“When making anomaly detection on the test data, CMAnomaly can be less sensitive and only raise the anomaly score when a real anomaly happens, avoiding more false alarms than USAD.”
BackgroundMethodInfluentialAn Unsupervised Long- and Short-Term Sparse Graph Neural Network for Multisensor Anomaly Detection
Miao, Q., Wang, D., Xu, C., Zhan, J., & Wu, C. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3383665 · Cited by 16
“2) Threshold Search: The threshold search method follows previous advanced research and uses the best threshold selection [13], [21], [22].”
MethodChen, F., Zhang, Y., Qin, Z., Fan, L., Jiang, R., Liang, Y., Wen, Q., & Deng, S. (2024) · IEEE International Conference on Data Engineering (ICDE) · doi:10.1109/icde60146.2024.00047 · Cited by 27
Zhao, X., Liu, P., Mahmoudi, S., Garg, S., Kaddoum, G., & Hassan, M. M. (2024) · Alexandria Engineering Journal · doi:10.1016/j.aej.2024.07.013 · Cited by 14
Zhou, G., Zheng, S., Yang, S., & Yi, S. (2024) · Science and Technology of Nuclear Installations · doi:10.1155/stni/9455897 · Cited by 10
Haq, I. U., Lee, B. S., & Rizzo, D. M. (2024) · Neural Computing and Applications · doi:10.1007/s00521-024-10759-1 · Cited by 18
Yao, M., Tao, D., Qi, P., & Gao, R. (2025) · IEEE Transactions on Services Computing · doi:10.1109/tsc.2025.3536306 · Cited by 12
“Ontheotherhand,someworks deployanadversarialtrainingprotocoltoforcethedecoderto learnamorerobustrepresentationtranslationapproach[11].”
Empowered Cyber–Physical Systems security using both network and physical data
Canonico, R., Esposito, G., Navarro, A., Romano, S. P., Sperlí, G., & Vignali, A. (2025) · Computers & Security · doi:10.1016/j.cose.2025.104382 · Cited by 12
Li, Z., Guo, W., An, J., Wang, Q., Mei, Y., Juan, R., Wang, T., Li, Y., & Gao, Z. (2025) · IEEE Sensors Journal · doi:10.1109/jsen.2025.3543835 · Cited by 13
“…superiority, we selected SOTA anomaly detection methods for comparison, including Clustering-based method: DeepSVDD [35]; Reconstruction-based methods: USAD [20], TranAD [36], and AMFormer [37]; Prediction-based methods: GDN [38] and GTA [23]; Discrepancy-based methods: AT [24] and DCDetector [34].”
Lu, R., Yang, Q., Zheng, D., Cao, W., Chen, X., & Li, C. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3546365 · Cited by 11
“and several state-of-art models, including OCSVM, kNN, iForest, LOF, kernel density estimation (KDE), AE, MemAE, deep autoencoding Gaussian mixture model (DAGMM) [27], unsupervised AD (USAD) [28], and variational AE-long short-term memory (VAE-LSTM) [29].”
TiTAD: Time-Invariant Transformer for Multivariate Time Series Anomaly Detection
Liu, Y., Wang, W., & Wu, Y. (2025) · Electronics · doi:10.3390/electronics14071401 · Cited by 13
Switching Gaussian Mixture Variational RNN for Anomaly Detection of Diverse CDN Websites
Dai, L., Chen, W., Liu, Y., Argyriou, A., Liu, C., Lin, T., Wang, P., Xu, Z., & Chen, B. (2022) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom48880.2022.9796836 · Cited by 15
“Among them, one line of the research mainly focus on learning of the spatial characteristics in the multivariate metrics but ignore the temporal dependency across the varying timesteps [4], [6], [31].”
BackgroundA Multivariate KPIs Anomaly Detection Framework With Dynamic Balancing Loss Training
Zhao, N., Han, B., Li, R., Su, J., & Zhou, C. (2022) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2022.3224803 · Cited by 11
“USAD [16] is an unsupervised method for detecting multivariate time-series anomalies.”
BackgroundMethodInfluentialAnomaly Detection in Cyber-Physical Systems based on BiGRU-VAE
Alguliyev, R. M., Sukhostat, L. V., & Mammadov, A. (2022) · IEEE International Conference on Application of Information and Communication Technologies (AICT) · doi:10.1109/aict55583.2022.10013581 · Cited by 13
ML Models for Detecting QoE Degradation in Low-Latency Applications: A Cloud-Gaming Case Study
Ky, J. R., Mathieu, B., Lahmadi, A., & Boutaba, R. (2023) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2023.3293806 · Cited by 15
“This class of algorithms, include Principal Component Analysis (PCA) and neural network algorithms such as AutoEncoder, LSTM-VAE [28], DAGMM [4], OmniAnomaly [29], Donut [9], and USAD [5].”
BackgroundMethodResultInfluentialVariable-wise generative adversarial transformer in multivariate time series anomaly detection
Yang, X., Li, H., Feng, X., & Jin, Z. (2023) · Applied Intelligence · doi:10.1007/s10489-023-05029-x · Cited by 11
Tuli, S., Casale, G., & Jennings, N. R. (2023) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2023.3330679 · Cited by 12
“To model the dependence of a data point x t at a timestamp t , we consider a sliding window of length k , denoted as as routinely done in prior work [33].”
MethodReal-Time Synchronization in Neural Networks for Multivariate Time Series Anomaly Detection
Abdulaal, A., & Lancewicki, T. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9413847 · Cited by 11
“For scoring results, we use a point-adjusted technique in line with recent work in literature [23]; If an anomaly was detected in the first 15 minutes of a site problem, then the full length of the problem is marked as true-positive.”
MethodLi, Q., Ji, Y., Sun, L., Zhang, N., & Yang, T. (2026) · Computers in Industry · doi:10.1016/j.compind.2026.104445 · Cited by 17
An Empirical Analysis of Anomaly Detection Methods for Multivariate Time Series
Li, D., Zhang, S., Sun, Y., Guo, Y., Che, Z., Chen, S., Zhong, Z., Liang, M., Shao, M., Li, M., et al. (2023) · Proceedings/Proceedings - International Symposium on Software Reliability Engineering · doi:10.1109/issre59848.2023.00014 · Cited by 13
“Several notable models in this category include MSCRED [6], DAGMM [7], USAD [12], DOMI [9], OmniAnomaly [13], SDFVAE [14], and InterFusion [16], which utilize an autoencoder (AE) or a variational AE (VAE) as their underlying framework.”
MethodHuang, X., Zhang, F., Fan, H., Chang, H., Zhou, B., & Li, Z. (2024) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2024.107928 · Cited by 11
HFN: Heterogeneous feature network for multivariate time series anomaly detection
Zhan, J., Wu, C., Yang, C., Miao, Q., & Ma, X. (2024) · Information Sciences · doi:10.1016/j.ins.2024.120626 · Cited by 14
Correlation-driven multi-level learning for anomaly detection on multiple energy sources
Kim, T., Jang, J.-S., & Kwon, H.-Y. (2024) · Applied Soft Computing · doi:10.1016/j.asoc.2024.111636 · Cited by 14
He, S., Li, G., Wang, J., Xie, K., & Sharma, P. K. (2024) · International Journal of Machine Learning and Cybernetics · doi:10.1007/s13042-024-02212-5 · Cited by 10
Ahmad, R., Alhasan, W., Wazirali, R., & Almajalid, R. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3407158 · Cited by 12
“However, unsupervised anomaly detection methods [11] have emerged as increasingly desirable in various fields due to their versatility and adaptability [28].”
BackgroundAnomaly Detection in Event-Triggered Traffic Time Series via Similarity Learning
Dou, S., Yang, K., Jiao, Y., Qiu, C., & Ren, K. (2024) · IEEE Transactions on Dependable and Secure Computing · doi:10.1109/tdsc.2024.3418906 · Cited by 16
“2) Baselines: For anomaly detection, we compare ET-Net against the following state-of-the-art unsupervised methods, including One-Class SVM (OCSVM), Local Outlier Factor (LoF), Isolation Forest (IF), Dynamic Time Warping (DTW), Kit-NET [58], GRU-AE [59], Shared-SRNN [49], DAGMM [28], and USAD [60].”
Multivariate Time Series Anomaly Detection: Fancy Algorithms and Flawed Evaluation Methodology
Sehili, M. E. A., & Zhang, Z. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-68031-1_1 · Cited by 16
DeepCache: Revisiting Cache Side-Channel Attacks in Deep Neural Networks Executables
Liu, Z., Yuan, Y., Chen, Y., Hu, S., Li, T., & Wang, S. (2024) · ACM SIGSAC Conference on Computer and Communications Security (CCS) · doi:10.1145/3658644.3690241 · Cited by 22
UltraADV: An Unsupervised Deep Learning Lightweight Framework for Anomaly Detection in V2X
Huang, W., Xu, H., Gong, Y., Liu, Z., Li, F., Lin, Z.-H., & Hu, B. (2024) · IEEE Internet of Things Journal · doi:10.1109/jiot.2024.3522581 · Cited by 14
“Deep learning has great potential in anomaly detection [12], [13], [14], [15], [16].”
He, Y., Ding, X. L., Tang, Y., Guan, J., & Zhou, S. (2025) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2025.3533437 · Cited by 11
“The five unsupervised MTS-AD methods include three reconstruction-based models: LSTM-AE [17], GDN [24], and USAD [52], a clustering-based method DeepSVDD [13], and a transformer-based model TranAD [21].”
Frequency-enhanced and decomposed transformer for multivariate time series anomaly detection
Li, S., Wang, Z., Wang, X., Yin, Z., & Yao, M. (2025) · Applied Intelligence · doi:10.1007/s10489-025-06441-1 · Cited by 15
He, S., He, W., Du, M., Jiang, X., & Dong, Y. (2025) · Information Sciences · doi:10.1016/j.ins.2025.122790 · Cited by 16
Zhang, G., Gao, X., Wang, L., Xue, B., Fu, S., Yu, J., Huang, Z., & Huang, X. (2022) · Applied Intelligence · doi:10.1007/s10489-022-04324-3 · Cited by 11
Deep Federated Anomaly Detection for Multivariate Time Series Data
Zhu, W., Song, D., Chen, Y., Cheng, W. D., Zong, B., Mizoguchi, T., Lumezanu, C., Chen, H., & Luo, J. (2022) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata55660.2022.10064694 · Cited by 12
“• USAD [21] is based on autoencoder framework.”
BackgroundMethodInfluentialFornás, J. G., Herrero, E., & Llombart, A. (2023) · IEEE Access · doi:10.1109/access.2023.3316600 · Cited by 11
“The article [22] presents USAD (Unsupervised Anomaly Detection for Multivariate Time Series), a fast and stable method based on adversarial trained autoencoders.”
MethodBeyond Sharing: Conflict-Aware Multivariate Time Series Anomaly Detection
Si, H., Pei, C., Li, Z., Zhao, Y., Li, J., Zhang, H., Diao, Z., Li, J., Xie, G., & Pei, D. (2023) · ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE) · doi:10.1145/3611643.3613896 · Cited by 21
“We visualize a few metrics in machine-1-8 included in SMD, on which models’ performance varies significantly, from 0.5761 to 0.9782 (Fig.”
BackgroundMethodInfluentialUnivariate Time Series Anomaly Detection Based on Hierarchical Attention Network
Chen, Z., Jia, D., Sun, Y., Yang, L., Jin, W., & Liu, R. (2024) · Tsinghua Science & Technology · doi:10.26599/tst.2023.9010073 · Cited by 12
Exploiting negative correlation for unsupervised anomaly detection in contaminated time series
Lin, X., Li, Z., Fan, H., Fu, Y.-G., & Chen, X. (2024) · Expert Systems with Applications · doi:10.1016/j.eswa.2024.123535 · Cited by 12
Frequency Domain Deep Learning With Non-Invasive Features for Intraoperative Hypotension Prediction
Moon, J.-H., Lee, G., Lee, S. M., Ryu, J., Kim, D., & Sohn, K.-A. (2024) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2024.3403109 · Cited by 18
“Specifically, we have integrated both classic deep learning models such as LSTM and VAE and a GAN-based model of USAD [25], as well as anomaly detection models based on Transformers such as Anomaly Transformer [26] and MEMTO [27].”
MethodAn Efficient Method for Detecting Abnormal Electricity Behavior
Tang, C., Qin, Y., Liu, Y., Pi, H., & Tang, Z. (2024) · Energies · doi:10.3390/en17112502 · Cited by 9
Model based anomaly detection in cyber physical power systems
Awad, A. M., Alheeti, K. M. A., & Najem, A. K. A. H. (2024) · AIP conference proceedings · doi:10.1063/5.0234148 · Cited by 9
Gao, Y. Q., Qi, J., Ye, H., Sun, Y., Hu, X., Dong, Z., & Sun, Y. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3502890 · Cited by 10
A Robust Framework for Evaluation of Unsupervised Time-Series Anomaly Detection
Güngör, O., Rios, A., Mudgal, P., Ahuja, N., & Rosing, T. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-78395-1_4 · Cited by 16
Combining KNN with AutoEncoder for Outlier Detection
Liu, S., Ma, S., Chen, H.-Q., Cui, L., & Ding, J. (2024) · Journal of Computer Science and Technology · doi:10.1007/s11390-023-2403-y · Cited by 11
Jiang, B., Wang, G., Cui, X., Luo, F., & Wang, J. (2025) · Internet of Things · doi:10.1016/j.iot.2025.101518 · Cited by 9
Huang, Y., Liu, P., Han, R., Yin, T., Dou, Q., Song, Y., & Luo, M. (2025) · IEEE Access · doi:10.1109/access.2025.3557571 · Cited by 12
Zhang, Z., Han, Y. M., & Geng, Z. (2025) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2025.3568019 · Cited by 13
“…the MoPIN is compared with ten baselines, including tranditional methods with the KPCA [14], the IF [5], the COPOD [16], the ECOD [17], and deep learning-based methods with the DAGMM [18], the OmniAnomaly [34], the USAD [19], the CAE M [35], the DCDetector [4], the TimesNet [20], and the ImAD [23].”
Self-Supervised ECG Anomaly Detection Based on Time-Frequency Specific Waveform Mask Feature Fusion
Tian, C., & Zhang, F. (2025) · IEEE Access · doi:10.1109/access.2025.3572484 · Cited by 10
“USAD [34] adopts a two-phase adversarially trained autoencoder, utilizing an encoder and two decoders to enhance adversarial learning and improve anomaly detection.”
Sánchez-Ferrera, A., Calvo, B., & Lozano, J. A. (2025) · ACM Computing Surveys · doi:10.1145/3770575 · Cited by 27
You, P., Chen, P., Li, X., & Bian, A. (2026) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2026.114104 · Cited by 12
Are we certain it’s anomalous?
Flaborea, A., Prenkaj, B., Munjal, B., Sterpa, M. A., d’Aragona, D. A., Podo, L., & Galasso, F. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00291 · Cited by 18
A Hybrid P4/NFV Architecture for Cloud Gaming Traffic Detection with Unsupervised ML
Ky, J. R., Graff, P., Mathieu, B., & Cholez, T. (2023) · IEEE Symposium on Computers and Communications (ISCC) · doi:10.1109/iscc58397.2023.10217863 · Cited by 14
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.