TY - GEN
T1 - Forecasting Error Pattern-Based Anomaly Detection in Multivariate Time Series
AU - Park, Seoyoung
AU - Han, Siho
AU - Woo, Simon S.
N1 - Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
PY - 2021
Y1 - 2021
N2 - The advent of Industry 4.0, partly characterized by the development of cyber-physical systems (CPSs), naturally entails the need for reliable security schemes. In particular, accurate detection of anomalies is of paramount importance, as even a small number of anomalous instances can trigger a catastrophic failure, often leading to a cascading one, throughout the CPS due to its interconnectivity. In this work, we aim to contribute to the body of literature on the application of anomaly detection techniques in CPSs. We propose novel Functional Data Analysis (FDA) and Autoencoder-based approaches for anomaly detection in the Secure Water Treatment (SWaT) dataset, which realistically represents a scaled-down industrial water treatment plant. We demonstrate that our methods can capture the underlying forecasting error patterns of the SWaT dataset generated by Mixture Density Networks (MDNs). We evaluate our detection performances using the F1 score and show that our methods empirically outperform the baseline approaches—cumulative sum (CUSUM) and static thresholding. We also provide a comparative analysis of our methods to discuss their abilities as well as limitations.
AB - The advent of Industry 4.0, partly characterized by the development of cyber-physical systems (CPSs), naturally entails the need for reliable security schemes. In particular, accurate detection of anomalies is of paramount importance, as even a small number of anomalous instances can trigger a catastrophic failure, often leading to a cascading one, throughout the CPS due to its interconnectivity. In this work, we aim to contribute to the body of literature on the application of anomaly detection techniques in CPSs. We propose novel Functional Data Analysis (FDA) and Autoencoder-based approaches for anomaly detection in the Secure Water Treatment (SWaT) dataset, which realistically represents a scaled-down industrial water treatment plant. We demonstrate that our methods can capture the underlying forecasting error patterns of the SWaT dataset generated by Mixture Density Networks (MDNs). We evaluate our detection performances using the F1 score and show that our methods empirically outperform the baseline approaches—cumulative sum (CUSUM) and static thresholding. We also provide a comparative analysis of our methods to discuss their abilities as well as limitations.
KW - Anomaly detection
KW - Cyber-physical systems
KW - Forecasting error patterns
UR - https://www.scopus.com/pages/publications/85103250001
U2 - 10.1007/978-3-030-67667-4_10
DO - 10.1007/978-3-030-67667-4_10
M3 - Conference contribution
AN - SCOPUS:85103250001
SN - 9783030676667
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 157
EP - 172
BT - Machine Learning and Knowledge Discovery in Databases
A2 - Dong, Yuxiao
A2 - Mladenic, Dunja
A2 - Saunders, Craig
PB - Springer Science and Business Media Deutschland GmbH
T2 - European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2020
Y2 - 14 September 2020 through 18 September 2020
ER -
