Bank of models: sensor attack detection and isolation in industrial control systems

Chuadhry Mujeeb Ahmed, Jianying Zhou

Research output: Contribution to conferencePaperpeer-review

Abstract

Attacks on sensor measurements can take the system to an unwanted state. The disadvantage of using a system model-based approach for attack detection is that it could not isolate which sensor was under attack. For example, if one of two sensors that are physically coupled is under attack, the attack would reflect in both. In this work, we propose an attack detection and isolation technique using a multi-model framework named Bank of Models (BoM) in which the same process will be represented by multiple system models. This technique can achieve higher accuracy for attack detection with low false alarm rates. We make extensive empirical performance evaluation on a realistic ICS testbed to demonstrate the viability of this technique.
Original languageEnglish
Number of pages20
Publication statusPublished - 29 Sep 2021
EventThe 16th International Conference on Critical Information Infrastructures Security - EPFL SwissTech Convention Center, Lausanne, Switzerland
Duration: 27 Sep 202129 Sep 2021
Conference number: 16th
https://critis2021.org/

Conference

ConferenceThe 16th International Conference on Critical Information Infrastructures Security
Abbreviated titleCRITIS 2021
Country/TerritorySwitzerland
CityLausanne
Period27/09/2129/09/21
Internet address

Keywords

  • sensor security
  • attack detection
  • industrial control systems

Fingerprint

Dive into the research topics of 'Bank of models: sensor attack detection and isolation in industrial control systems'. Together they form a unique fingerprint.

Cite this