Back to search

University of Patras

Αντίπαλες τακτικές ελέγχου για κυβερνοφυσικά συστήματα

Abstract

dc:description

This dissertation addresses the design of attack scenarios along with the development of suitable attack detection methods using control-theoretic tools, thus giving rise to adversary control tactics in cyber-physical systems. The main characteristic of the systems under consideration is the existence of operational constraints, which are associated with both the control input and the state vector. The constraints are polyhedral in nature and represent saturation bounds, when it comes to the control input, and safety or performance criteria, when it comes to the state vector. We demonstrate that these constraints can be exploited for the design of effective attacks using tools from the classical control theory, such as the polyhedral Lyapunov functions, in conjunction with modern optimization techniques. Furthermore, it is shown that based on the same constraints we can develop enhanced attack detection methods with the help of set-theoretic tools, such as the robust invariant sets. The assessment of our research results is performed in terms of the analysis of a cyber-physical electric power system and more specifically through the study of the load-frequency control loop. The transmission of measurements from remote sensors through unprotected communication channels along with the absence of any human-operators in the various data management layers justify completely the selection of this particular control loop for the purposes of a cybersecure-oriented study. The model that we used to describe the interactions between the individual control areas of a generic electric power network is linear and time invariant. Furthermore, we chose to represent the evolution of the dynamic phenomena in the discrete-time domain in order to reduce the computational cost during the design of the attacks and the detectors. The cases of an islanded and a networked power system, both in the presence and in the absence of a power load, where examined separately. The dissertation is focused on the study of two attack scenarios. Initially, we study the scenario, where the system is denied service from the control center and the input signal is supplanted by the attack signal, whereas in the sequel we study the scenario where the measurement data, which are collected by the sensors and used by the system controllers, are corrupted. In both cases, the aim of the attacker is to cause system malfunctions, either by deregulating the electrical frequency from its nominal value or by desynchronizing the coupled generators. Special effort was made to design effective attacks, while gradually reducing the knowledge and information resources required about the system characteristics, with the purpose of studying realistic scenarios. We addressed the cases of both persistent and switching attacks. For the design of the detectors, a set-theoretic approach was adopted. The key idea is the activation of an alarm signal, whenever the state vector trajectory exits a robust invariant polyhedral set. The robustness property provides us with the ability to detect attacks even in the case where the system is simultaneously affected by power load changes, assuming that the latter ones remain bounded. The design of the detector is complemented by an alternative modeling method of the networked power system, which prompts certain state variables to behave in an unstable manner, when the measurement data are corrupted. It is shown that the convex and compact nature of the robust invariant sets along with the unstable system behavior, in the presence of an attacker, guarantee the detection of both persistent and switching data corruption attack patterns, something that the conventional detectors, which are implemented as state estimators, are unable to ensure.

Degree

thesis:*
Grantor dc:publisher
University of Patras
Year dc:date
2020

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Kontouras, Efstathios
  • Κοντουράς, Ευστάθιος

Subjects

dc:subject × 17

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/47370

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Kontouras, Efstathios; Κοντουράς, Ευστάθιος. Αντίπαλες τακτικές ελέγχου για κυβερνοφυσικά συστήματα. University of Patras, 2020. http://hdl.handle.net/10442/hedi/47370