Skip to main navigation Skip to search Skip to main content

Cooperative Attack Detection of Power CPS based on Feature Relation Graph Convolutional Network

  • Beihang University

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

Abstract

Cooperative attack is a main threat to power cyber-physical system, which has a high success rate and strong destructiveness. The existing detection methods are limited by a single data type, so it is difficult to detect the combination of cooperative attacks and take defensive measures. In this paper, we propose feature relation graph convolutional network for cooperative attack detection by means of extracting the feature relationship and power node topology relationship. In this method, the power monitoring system data and the communication network data are fused to improve the efficiency of cooperative attack detection. Compared with graph convolutional network, the training time of the proposed method is reduced by 67.78%-94.60% and the detection accuracy is improved by 0.65%-3.41%.

Original languageEnglish
Title of host publicationProceedings - 2022 IEEE 19th International Conference on Mobile Ad Hoc and Smart Systems, MASS 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages380-384
Number of pages5
ISBN (Electronic)9781665471800
DOIs
StatePublished - 2022
Event19th IEEE International Conference on Mobile Ad Hoc and Smart Systems, MASS 2022 - Denver, United States
Duration: 20 Oct 202222 Oct 2022

Publication series

NameProceedings - 2022 IEEE 19th International Conference on Mobile Ad Hoc and Smart Systems, MASS 2022

Conference

Conference19th IEEE International Conference on Mobile Ad Hoc and Smart Systems, MASS 2022
Country/TerritoryUnited States
CityDenver
Period20/10/2222/10/22

Keywords

  • Attack Detection
  • Graph Convolutional Network
  • Power Cyber-Physical System

Fingerprint

Dive into the research topics of 'Cooperative Attack Detection of Power CPS based on Feature Relation Graph Convolutional Network'. Together they form a unique fingerprint.

Cite this