Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders

Qiwen Li; Tijin Yan; Huanhuan Yuan; Yuanqing Xia

doi:10.23919/CCC55666.2022.9902551

Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders

Qiwen Li, Tijin Yan, Huanhuan Yuan, Yuanqing Xia^*

^*Corresponding author for this work

School of Automation

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Data-driven anomaly detection continues to be challenging due to the increased complexity of modern cyber physical systems (CPSs) and their temporal dependencies. Unsupervised detection techniques are widely used through VAE-based frame-works and RNN-based deep learning techniques. However, VAE and its variants impose too much constraint on extracted latent code, and RNN's autoregressive essence indicates the shortage of parallelism and long-term prediction. To tackle the above is-sues, we propose TransAAE (Transformer-augmented Adversarial Autoencoder), a novel unsupervised approach for multivariate time series anomaly detection. The use of the adversarial autoencoder (AAE) architecture loosens the regularization of latent code, and self-attention mechanism is utilized to extract temporal information. Extensive experiments show an average F1 score over 0.9 on three public datasets, which significantly outperforms among the baselines.

Original language	English
Title of host publication	Proceedings of the 41st Chinese Control Conference, CCC 2022
Editors	Zhijun Li, Jian Sun
Publisher	IEEE Computer Society
Pages	4251-4256
Number of pages	6
ISBN (Electronic)	9789887581536
DOIs	https://doi.org/10.23919/CCC55666.2022.9902551
Publication status	Published - 2022
Event	41st Chinese Control Conference, CCC 2022 - Hefei, China Duration: 25 Jul 2022 → 27 Jul 2022

Publication series

Name	Chinese Control Conference, CCC
Volume	2022-July
ISSN (Print)	1934-1768
ISSN (Electronic)	2161-2927

Conference

Conference	41st Chinese Control Conference, CCC 2022
Country/Territory	China
City	Hefei
Period	25/07/22 → 27/07/22

Keywords

Adversarial Autoencoder
Anomaly Detection
Cyber Physical Systems
Self-Attention Mechanism

Access to Document

10.23919/CCC55666.2022.9902551

Cite this

Li, Q., Yan, T., Yuan, H., & Xia, Y. (2022). Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders. In Z. Li, & J. Sun (Eds.), Proceedings of the 41st Chinese Control Conference, CCC 2022 (pp. 4251-4256). (Chinese Control Conference, CCC; Vol. 2022-July). IEEE Computer Society. https://doi.org/10.23919/CCC55666.2022.9902551

@inproceedings{bd3ba62c01ed4bbf961c6c4ce8628cac,

title = "Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders",

abstract = "Data-driven anomaly detection continues to be challenging due to the increased complexity of modern cyber physical systems (CPSs) and their temporal dependencies. Unsupervised detection techniques are widely used through VAE-based frame-works and RNN-based deep learning techniques. However, VAE and its variants impose too much constraint on extracted latent code, and RNN's autoregressive essence indicates the shortage of parallelism and long-term prediction. To tackle the above is-sues, we propose TransAAE (Transformer-augmented Adversarial Autoencoder), a novel unsupervised approach for multivariate time series anomaly detection. The use of the adversarial autoencoder (AAE) architecture loosens the regularization of latent code, and self-attention mechanism is utilized to extract temporal information. Extensive experiments show an average F1 score over 0.9 on three public datasets, which significantly outperforms among the baselines.",

keywords = "Adversarial Autoencoder, Anomaly Detection, Cyber Physical Systems, Self-Attention Mechanism",

author = "Qiwen Li and Tijin Yan and Huanhuan Yuan and Yuanqing Xia",

note = "Publisher Copyright: {\textcopyright} 2022 Technical Committee on Control Theory, Chinese Association of Automation.; 41st Chinese Control Conference, CCC 2022 ; Conference date: 25-07-2022 Through 27-07-2022",

year = "2022",

doi = "10.23919/CCC55666.2022.9902551",

language = "English",

series = "Chinese Control Conference, CCC",

publisher = "IEEE Computer Society",

pages = "4251--4256",

editor = "Zhijun Li and Jian Sun",

booktitle = "Proceedings of the 41st Chinese Control Conference, CCC 2022",

address = "United States",

}

Li, Q, Yan, T, Yuan, H & Xia, Y 2022, Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders. in Z Li & J Sun (eds), Proceedings of the 41st Chinese Control Conference, CCC 2022. Chinese Control Conference, CCC, vol. 2022-July, IEEE Computer Society, pp. 4251-4256, 41st Chinese Control Conference, CCC 2022, Hefei, China, 25/07/22. https://doi.org/10.23919/CCC55666.2022.9902551

Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders. / Li, Qiwen; Yan, Tijin; Yuan, Huanhuan et al.
Proceedings of the 41st Chinese Control Conference, CCC 2022. ed. / Zhijun Li; Jian Sun. IEEE Computer Society, 2022. p. 4251-4256 (Chinese Control Conference, CCC; Vol. 2022-July).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders

AU - Li, Qiwen

AU - Yan, Tijin

AU - Yuan, Huanhuan

AU - Xia, Yuanqing

PY - 2022

Y1 - 2022

N2 - Data-driven anomaly detection continues to be challenging due to the increased complexity of modern cyber physical systems (CPSs) and their temporal dependencies. Unsupervised detection techniques are widely used through VAE-based frame-works and RNN-based deep learning techniques. However, VAE and its variants impose too much constraint on extracted latent code, and RNN's autoregressive essence indicates the shortage of parallelism and long-term prediction. To tackle the above is-sues, we propose TransAAE (Transformer-augmented Adversarial Autoencoder), a novel unsupervised approach for multivariate time series anomaly detection. The use of the adversarial autoencoder (AAE) architecture loosens the regularization of latent code, and self-attention mechanism is utilized to extract temporal information. Extensive experiments show an average F1 score over 0.9 on three public datasets, which significantly outperforms among the baselines.

AB - Data-driven anomaly detection continues to be challenging due to the increased complexity of modern cyber physical systems (CPSs) and their temporal dependencies. Unsupervised detection techniques are widely used through VAE-based frame-works and RNN-based deep learning techniques. However, VAE and its variants impose too much constraint on extracted latent code, and RNN's autoregressive essence indicates the shortage of parallelism and long-term prediction. To tackle the above is-sues, we propose TransAAE (Transformer-augmented Adversarial Autoencoder), a novel unsupervised approach for multivariate time series anomaly detection. The use of the adversarial autoencoder (AAE) architecture loosens the regularization of latent code, and self-attention mechanism is utilized to extract temporal information. Extensive experiments show an average F1 score over 0.9 on three public datasets, which significantly outperforms among the baselines.

KW - Adversarial Autoencoder

KW - Anomaly Detection

KW - Cyber Physical Systems

KW - Self-Attention Mechanism

UR - http://www.scopus.com/inward/record.url?scp=85140436570&partnerID=8YFLogxK

U2 - 10.23919/CCC55666.2022.9902551

DO - 10.23919/CCC55666.2022.9902551

M3 - Conference contribution

AN - SCOPUS:85140436570

T3 - Chinese Control Conference, CCC

SP - 4251

EP - 4256

BT - Proceedings of the 41st Chinese Control Conference, CCC 2022

A2 - Li, Zhijun

A2 - Sun, Jian

PB - IEEE Computer Society

T2 - 41st Chinese Control Conference, CCC 2022

Y2 - 25 July 2022 through 27 July 2022

ER -

Self-Attention-Based Multivariate Anomaly Detection for CPS Time Series Data with Adversarial Autoencoders

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this