USAD: UnSupervised Anomaly Detection on Multivariate Time Series

By Julien Audibert, Pietro Michiardi, Frédéric Guyard, Sébastien Marti, Maria A. Zuluaga. Knowledge Discovery and Data Mining, 2020

1,210 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Computer Networks and Communications, Signal Processing

View Paper: https://www.semanticscholar.org/paper/7c7fc1b774e4cf265de151ed4cf63e3266a978be

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Jules Samaran
2. Astha Garg
3. Dan Pei
4. Theodoros Tsiligkaridis
5. Jihun Yi
6. Jiaqi Han
7. Rui Jiao
8. Marinka Zitnik

In Computer Networks and Communications

1. Jihun Yi
2. Marinka Zitnik
3. Kukjin Choi
4. Ahmed Abdulaal
5. Tomer Lancewicki
6. Sungroh Yoon
7. Changhwa Park
8. Zhuanghua Liu

In Signal Processing

1. Jihun Yi
2. Marinka Zitnik
3. Chuan-Sheng Foo
4. Kukjin Choi
5. Ahmed Abdulaal
6. Tomer Lancewicki
7. Sungroh Yoon
8. Changhwa Park

Citations in the corpus

• An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series (2021, methodology): “score even when a large number of events are not detected [8].”
• Deep Learning for Anomaly Detection in Time-Series Data: Review, Analysis, and Guidelines (2021, methodology): “Audibert et al. [72] proposed AE, in which one encoder and two decoders are trained adversarially, to identify network anomalies.”
• Multivariate Time Series Anomaly Detection and Interpretation using Hierarchical Inter-Metric and Temporal Embedding (2021, methodology): “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].”
• Practical Approach to Asynchronous Multivariate Time Series Anomaly Detection and Localization (2021, methodology): “Alternatively, USAD is based on adversarial learning.”
• Self-Supervised Contrastive Pre-Training For Time Series via Time-Frequency Consistency (2022, background): “Based on global patterns, the former aims to detect abnormal time series samples instead of outlier observations in a sample (as in BTSF [50] and USAD [70]) which emphasizes local context.”
• Deep Learning for Time Series Anomaly Detection: A Survey (2022, neutral): “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.”