Towards an Awareness of Time Series Anomaly Detection Models' Adversarial Vulnerability

被引:4
作者
Tariq, Shahroz [1 ]
Le, Binh M. [2 ]
Woo, Simon S. [3 ]
机构
[1] Data61 CSIRO, Sydney, NSW, Australia
[2] Sungkyunkwan Univ, Coll Comp & Informat, Seoul, South Korea
[3] Sungkyunkwan Univ, Dept Artificial Intelligence, Seoul, South Korea
来源
PROCEEDINGS OF THE 31ST ACM INTERNATIONAL CONFERENCE ON INFORMATION AND KNOWLEDGE MANAGEMENT, CIKM 2022 | 2022年
基金
新加坡国家研究基金会;
关键词
Adversarial Attack; Anomaly Detection; Time Series; Classification;
D O I
10.1145/3511808.3557073
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Time series anomaly detection is extensively studied in statistics, economics, and computer science. Over the years, numerous methods have been proposed for time series anomaly detection using deep learning-based methods. Many of these methods demonstrate state-of-the-art performance on benchmark datasets, giving the false impression that these systems are robust and deployable in many practical and industrial real-world scenarios. In this paper, we demonstrate that the performance of state-of-the-art anomaly detection methods is degraded substantially by adding only small adversarial perturbations to the sensor data. We use different scoring metrics such as prediction errors, anomaly, and classification scores over several public and private datasets ranging from aerospace applications, server machines, to cyber-physical systems in power plants. Under well-known adversarial attacks from Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) methods, we demonstrate that state-of-the-art deep neural networks (DNNs) and graph neural networks (GNNs) methods, which claim to be robust against anomalies and have been possibly integrated in real-life systems, have their performance drop to as low as 0%. To the best of our understanding, we demonstrate, for the first time, the vulnerabilities of anomaly detection systems against adversarial attacks. The overarching goal of this research is to raise awareness towards the adversarial vulnerabilities of time series anomaly detectors.
引用
收藏
页码:3534 / 3544
页数:11
相关论文
共 50 条
[31]   Time Series Anomaly Detection using Diffusion-based Models [J].
Pintilie, Ioana ;
Manolache, Andrei ;
Brad, Florin .
2023 23RD IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS, ICDMW 2023, 2023, :570-578
[32]   Anomaly Detection at Scale: The Case for Deep Distributional Time Series Models [J].
Ayed, Fadhel ;
Stella, Lorenzo ;
Januschowski, Tim ;
Gasthaus, Jan .
SERVICE-ORIENTED COMPUTING, ICSOC 2020, 2021, 12632 :97-109
[33]   Embedding Models for Multivariate Time Series Anomaly Detection in Industry 5.0 [J].
Colombi, Lorenzo ;
Vespa, Michela ;
Belletti, Nicolas ;
Brina, Matteo ;
Dahdal, Simon ;
Tabanelli, Filippo ;
Resca, Francesco ;
Bellodi, Elena ;
Tortonesi, Mauro ;
Stefanelli, Cesare ;
Vignoli, Massimiliano .
DATA SCIENCE AND ENGINEERING, 2025,
[34]   A systematic literature review of IoT time series anomaly detection solutions [J].
Sgueglia, Arnaldo ;
Di Sorbo, Andrea ;
Visaggio, Corrado Aaron ;
Canfora, Gerardo .
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE, 2022, 134 :170-186
[35]   Robust Anomaly Detection of Multivariate Time Series Data via Adversarial Graph Attention BiGRU [J].
Xing, Yajing ;
Tan, Jinbiao ;
Zhang, Rui ;
Wan, Jiafu .
BIG DATA AND COGNITIVE COMPUTING, 2025, 9 (05)
[36]   Multivariate Time Series Anomaly Detection With Generative Adversarial Networks Based on Active Distortion Transformer [J].
Kong, Lingkun ;
Yu, Jinsong ;
Tang, Diyin ;
Song, Yue ;
Han, Danyang .
IEEE SENSORS JOURNAL, 2023, 23 (09) :9658-9668
[37]   Towards Deep Industrial Transfer Learning for Anomaly Detection on Time Series Data [J].
Maschler, Benjamin ;
Knodel, Tim ;
Weyrich, Michael .
2021 26TH IEEE INTERNATIONAL CONFERENCE ON EMERGING TECHNOLOGIES AND FACTORY AUTOMATION (ETFA), 2021,
[38]   Towards Unsupervised Time-Series Anomaly Detection for Virtual Cloud Networks [J].
Ma, Zixuan ;
Li, Chen ;
Zhang, Kun ;
Tu, Bibo .
IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY, 2025, 20 :4322-4337
[39]   MAD-GAN: Multivariate Anomaly Detection for Time Series Data with Generative Adversarial Networks [J].
Li, Dan ;
Chen, Dacheng ;
Shi, Lei ;
Jin, Baihong ;
Goh, Jonathan ;
Ng, See-Kiong .
ARTIFICIAL NEURAL NETWORKS AND MACHINE LEARNING - ICANN 2019: TEXT AND TIME SERIES, PT IV, 2019, 11730 :703-716
[40]   AdaDiffAD: Adaptively Segmenting Diffusion Models for Time Series Anomaly Detection in Dynamic JointCloud Environment [J].
Ma, Chao ;
Yi, Lin ;
Zhou, Linjiang ;
Wang, Zepeng ;
Shi, Xiaochuan ;
Zhu, Weiping .
2024 IEEE 30TH INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED SYSTEMS, ICPADS, 2024, :334-341