Towards Building Intrusion Detection Systems for Multivariate Time-Series Data

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Recent network intrusion detection systems have employed machine learning and deep learning algorithms to defend against dynamically evolving network attacks. While most previous studies have focused on detecting attacks which can be determined based on a single time instant, few studies have paid attention to subsequence outliers, which require inspecting consecutive points in time for detection. To address this issue, this paper applies a time-series anomaly detection method in an unsupervised learning manner. To this end, we converted the UNSW-NB15 dataset into the time-series data. We carried out a preliminary evaluation to test the performance of the anomaly detection on the created time-series network dataset as well as on a time-series dataset obtained from sensors. We analyze and discuss the results.

Original languageEnglish
Title of host publicationSilicon Valley Cybersecurity Conference
EditorsSang-Yoon Chang, Luis Bathen, Fabio Di Troia, Thomas H. Austin, Alex J. Nelson
PublisherSpringer Science and Business Media Deutschland GmbH
Pages45-56
Number of pages12
ISBN (Print)9783030960568
DOIs
StatePublished - 2022
Event2nd Silicon Valley Cybersecurity Conference, SVCC 2021 - Virtual, Online
Duration: 2 Dec 20213 Dec 2021

Publication series

NameCommunications in Computer and Information Science
Volume1536 CCIS
ISSN (Print)1865-0929
ISSN (Electronic)1865-0937

Conference

Conference2nd Silicon Valley Cybersecurity Conference, SVCC 2021
CityVirtual, Online
Period2/12/213/12/21

Keywords

  • Anomaly detection
  • Intrusion detection system
  • Stacked RNN
  • Time series
  • Unsupervised learning

Fingerprint

Dive into the research topics of 'Towards Building Intrusion Detection Systems for Multivariate Time-Series Data'. Together they form a unique fingerprint.

Cite this