IEEE Systems, Man and Cybernetics Magazine - April 2021 - 37

survey was on power systems and sensor networks. This
article provides an overview of secure control methods in
a wide range of CPSs.
The main contributions of this article are summarized
in the following.
◆◆ The article presents common terminologies, brief
descriptions, and the most significant CPS characteristics.
◆◆ Major categories of secure control strategies reported
in the open literature are reviewed, and there is a brief
discussion of their pros and cons.
◆◆ Notable CPS examples are outlined. In each domain,
special characteristics, underlying concepts, challenges, and research progress are discussed.
◆◆ Finally, based on the material reviewed in this article,
areas of potential future research are outlined.
The Most Significant CPS Characteristics
CPSs are distributed networks that employ SCADA to
monitor and control critical infrastructures [30]-[32].
They are integrated with the Internet, which creates an
open-access environment. In CPSs, physical systems and
software components are intertwined to provide an
intelligent structure. However, due to their availability
and accessibility, CPSs have a high risk of cyberattacks.
CPSs consist of several embedded sensors, actuators,
distributed controllers, SCADA, and, of course, a physical system to be controlled and monitored. Figure 1 represents the CPS structure. It is noted from Figure 1 that
the SCADA may communicate with several local controllers. Furthermore, the physical system is always distributed in a large area, with several subsystems. Intelligent
sensors cooperatively monitor the physical system and
measure environmental data, such as pressure, temperature, position, and speed, and report the information to
local controllers and SCADA. The SCADA monitors the
working condition of the entire system and sets operating points through the local controllers. The local controllers command the actuators that regulate outputs at
desired levels [33], [34].
While many physical systems operate in continuous
time, digital controllers are often designed and employed
because of advances in electronics and for practical implementation aspects. In designing digital controllers, continuous systems described by differential equations are often
transformed into discrete time systems described with difference equations. Given this, in what follows, we use difference equations to describe a discrete time multiple-input,
multiple-output dynamical system in the state-space form
	

x (k + 1) = f (x (k), u (k)) + w (k)

y (k) = g (x (k), u (k)) + v (k),

C1

C2

Local Controllers
SCADA

A1

Actuators

S2

A2

SS1

(1)

where x (k) = [x 1 (k), f, x n (k)]T ! R n is the state of the system at time k x and x is the sampling interval. Variable
u(k) = [u 1 (k), f, u i (k), f, u p (k)] ! R p is the input vector,
	

and y (k) = [y 1 (k), f, y i (k), f, y q (k)] ! R q describes the
output vector. Function f (x (k), u (k)) describes the mathematical representation of the nonlinear system and is a
nonlinear function of variables x(k) and u(k). Function
g (x (k), u(k)) is measurement model with a nonlinear
function of variables x(k) and u(k). Moreover, u i (k) and
y i (k) are the kth signal of input i and the kth measurement
observed by sensor i, respectively. Variable w (k) ! R n # 1
denotes process noise, which is a zero-mean white Gaussian sequence, whose variance is Q. Variable v(k) ! R q indicates measurement noise, which is a zero-mean white
Gaussian sequence whose variance is R. It is supposed that
the initial value of the state vector x(0) is a random vector
whose mean and estimation error covariance matrix are
x 0 and P0 , respectively. Furthermore, x(0), w(k), and v(k)
are mutually independent.
CPSs often consist of large, distributed physical subsystems that interact with one another as well as myriad software-based computational algorithms that provide
intelligence at many levels. This fact makes CPSs highly
complex engineering systems with certain significant characteristic requirements, such as availability, integrity, timeliness, and confidentiality. The following characteristics are
essential to CPSs because adversaries and hackers attempt
to breach one of them whenever they launch a malicious act.
◆◆ Availability: The availability requirement demands that
the CPSs be accessible and available at all times and
successfully fulfill what is expected of them [35]. Note
that the real-time availability requirement is stricter for
CPSs than it is for IT systems, where temporary
unavailability may have less-severe consequences [30].

Sensors
S2

SS2

Subsystems Physical System

Figure 1. A typical CPS.

Ap ri l 2021

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE	

37
IEEE Systems, Man and Cybernetics Magazine - April 2021

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - April 2021
