Robust detection of anomalies via sparse methods

Zoltán Milacski, Marvin Ludersdorfer, A. Lőrincz, Patrick Van Der Smagt

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

The problem of anomaly detection is a critical topic across application domains and is the subject of extensive research. Applic ationsinclude finding frauds and intrusions, warning on robot safety, and many others. Standard approaches in this field exploit simple or complex system models, created by experts using detailed domain knowledge. In this paper, we put forth a statistics-based anomaly detector motivated by the fact that anomalies are sparse by their very nature. Powerful sparsity directed algorithms—namely Robust Principal Component Analysis and the Group Fused LASSO—form the basis of the methodology. Our novel unsupervised single-step solution imposes a convex optimisation task on the vector time series data of the monitored system by employing group-structured, switching and robust regularisation techniques. We evaluated our method on data generated by using a Baxter robot arm that was disturbed randomly by a human operator. Our procedure was able to outperform two baseline schemes in terms of F1 score. Generalisations to more complex dynamical scenarios are desired.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages419-426
Number of pages8
Volume9491
ISBN (Print)9783319265544
DOIs
Publication statusPublished - 2015
Event22nd International Conference on Neural Information Processing, ICONIP 2015 - Istanbul, Turkey
Duration: Nov 9 2015Nov 12 2015

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9491
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other22nd International Conference on Neural Information Processing, ICONIP 2015
CountryTurkey
CityIstanbul
Period11/9/1511/12/15

Fingerprint

Anomaly
Robot
Robots
Regularization Technique
Convex optimization
Anomaly Detection
Domain Knowledge
Time Series Data
Convex Optimization
Sparsity
Principal component analysis
Principal Component Analysis
Large scale systems
Time series
Baseline
Complex Systems
Safety
Detector
Statistics
Detectors

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Milacski, Z., Ludersdorfer, M., Lőrincz, A., & Van Der Smagt, P. (2015). Robust detection of anomalies via sparse methods. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9491, pp. 419-426). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9491). Springer Verlag. https://doi.org/10.1007/978-3-319-26555-1_47

Robust detection of anomalies via sparse methods. / Milacski, Zoltán; Ludersdorfer, Marvin; Lőrincz, A.; Van Der Smagt, Patrick.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9491 Springer Verlag, 2015. p. 419-426 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9491).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Milacski, Z, Ludersdorfer, M, Lőrincz, A & Van Der Smagt, P 2015, Robust detection of anomalies via sparse methods. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 9491, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9491, Springer Verlag, pp. 419-426, 22nd International Conference on Neural Information Processing, ICONIP 2015, Istanbul, Turkey, 11/9/15. https://doi.org/10.1007/978-3-319-26555-1_47
Milacski Z, Ludersdorfer M, Lőrincz A, Van Der Smagt P. Robust detection of anomalies via sparse methods. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9491. Springer Verlag. 2015. p. 419-426. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-26555-1_47
Milacski, Zoltán ; Ludersdorfer, Marvin ; Lőrincz, A. ; Van Der Smagt, Patrick. / Robust detection of anomalies via sparse methods. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9491 Springer Verlag, 2015. pp. 419-426 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{e8955afab1d34b0d8f536bfba3421d3c,
title = "Robust detection of anomalies via sparse methods",
abstract = "The problem of anomaly detection is a critical topic across application domains and is the subject of extensive research. Applic ationsinclude finding frauds and intrusions, warning on robot safety, and many others. Standard approaches in this field exploit simple or complex system models, created by experts using detailed domain knowledge. In this paper, we put forth a statistics-based anomaly detector motivated by the fact that anomalies are sparse by their very nature. Powerful sparsity directed algorithms—namely Robust Principal Component Analysis and the Group Fused LASSO—form the basis of the methodology. Our novel unsupervised single-step solution imposes a convex optimisation task on the vector time series data of the monitored system by employing group-structured, switching and robust regularisation techniques. We evaluated our method on data generated by using a Baxter robot arm that was disturbed randomly by a human operator. Our procedure was able to outperform two baseline schemes in terms of F1 score. Generalisations to more complex dynamical scenarios are desired.",
author = "Zolt{\'a}n Milacski and Marvin Ludersdorfer and A. Lőrincz and {Van Der Smagt}, Patrick",
year = "2015",
doi = "10.1007/978-3-319-26555-1_47",
language = "English",
isbn = "9783319265544",
volume = "9491",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "419--426",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Robust detection of anomalies via sparse methods

AU - Milacski, Zoltán

AU - Ludersdorfer, Marvin

AU - Lőrincz, A.

AU - Van Der Smagt, Patrick

PY - 2015

Y1 - 2015

N2 - The problem of anomaly detection is a critical topic across application domains and is the subject of extensive research. Applic ationsinclude finding frauds and intrusions, warning on robot safety, and many others. Standard approaches in this field exploit simple or complex system models, created by experts using detailed domain knowledge. In this paper, we put forth a statistics-based anomaly detector motivated by the fact that anomalies are sparse by their very nature. Powerful sparsity directed algorithms—namely Robust Principal Component Analysis and the Group Fused LASSO—form the basis of the methodology. Our novel unsupervised single-step solution imposes a convex optimisation task on the vector time series data of the monitored system by employing group-structured, switching and robust regularisation techniques. We evaluated our method on data generated by using a Baxter robot arm that was disturbed randomly by a human operator. Our procedure was able to outperform two baseline schemes in terms of F1 score. Generalisations to more complex dynamical scenarios are desired.

AB - The problem of anomaly detection is a critical topic across application domains and is the subject of extensive research. Applic ationsinclude finding frauds and intrusions, warning on robot safety, and many others. Standard approaches in this field exploit simple or complex system models, created by experts using detailed domain knowledge. In this paper, we put forth a statistics-based anomaly detector motivated by the fact that anomalies are sparse by their very nature. Powerful sparsity directed algorithms—namely Robust Principal Component Analysis and the Group Fused LASSO—form the basis of the methodology. Our novel unsupervised single-step solution imposes a convex optimisation task on the vector time series data of the monitored system by employing group-structured, switching and robust regularisation techniques. We evaluated our method on data generated by using a Baxter robot arm that was disturbed randomly by a human operator. Our procedure was able to outperform two baseline schemes in terms of F1 score. Generalisations to more complex dynamical scenarios are desired.

UR - http://www.scopus.com/inward/record.url?scp=84952005783&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84952005783&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-26555-1_47

DO - 10.1007/978-3-319-26555-1_47

M3 - Conference contribution

AN - SCOPUS:84952005783

SN - 9783319265544

VL - 9491

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 419

EP - 426

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -