IEEE Electrification - September 2021 - 59

Here, we describe four attack templates that may
compromise the distribution system by manipulating
measurements.
x Replay attack: The attackers record a segment of true
measurements and replay the prerecording when
launching an attack. The replay attack blinds the
controllers and forces them to lose their regulation
capacities.
x Destabilization attack: The attackers aim to modify the
measurements for destabilizing the targeted system.
For example, suppose that an attacker attempts to
destabilize a microgrid by manipulating the frequency
measurement. Assume that the true frequency is
60.03 Hz and the secondary control is designed to
regulate the frequency to 60 Hz. Then, the attacker
forces the frequency sensor to report 59.97 Hz to
the MSn
. If this happens, the actual frequency may
further deviate from 60 Hz. The attacker can gradually
render the microgrid unstable by repeating
this procedure.
x Noise-injection attack: The attackers superpose noise
upon the true measurement(s) and force the compromised
sensors to report the measurements with large
noise to targeted controllers. The injected noise may
compromise energy efficiency.
x Harmonic-injection attack: The attackers superpose
fake harmonics upon the true measurement(s). These
fake harmonics could decrease the power factor, efficiency,
and power output of the inverter.
In the two-level P2P energy market described in the
last section, malicious market participants can financially
benefit from launching deliberated cyberattacks. This
increases the possibility that a cyberattack will occur. In
a distributed microgrid-level P2P market, participating
prosumers in a microgrid may have a competitive relationship.
Each prosumer maximizes his/her profits by
selling as much energy as possible with the highest
attainable price. Because there is competition among
prosumers, a rational prosumer may not set an arbitrarily
high price for trading energy since, if the energy price
is too high, a buyer may purchase energy from other prosumers.
However, in an extreme case, a malicious prosumer
may monopolize the microgrid market and set
the most favorable price by illegally launching cyberattacks
on his/her competitors. For example, the prosumer
may launch a destabilization attack on the inverters of
his/her possible competitors, making the competitors fail
to deliver energy. Similarly, in the distribution systemlevel
P2P market, the market participants are microgrid
stakeholders. It is also possible that a malicious
microgrid stakeholder could become a monopoly in the
P2P market by illegally compromising the system-level
control of other competitive microgrids. Therefore, a prerequisite
of establishing P2P energy markets that maximize
social welfare is to secure distribution systems
from malicious cyberattacks.
The Dynamic Watermarking Technique
This section introduces the basic principles of the
dynamic watermarking technique. Figure 6 recapitulates
the decision-making processes in microgrids. The
dynamic watermarking technique aims to secure the
dynamical system by performing two statistical tests on
sensor measurements of the system against arbitrary
cyberattacks. The key idea of the dynamic watermarking
technique is as follows. Instead of directly using the control
input command issued by the controller in Figure 6, a
small random private excitation signal ek6 @ is superimposed
on the control input. We define this ek6 @ as a
watermark because it creates a digital mark in a measurement
that cannot be removed in the system. Such a tiny
watermark signal possesses certain statistics that propagate
to the measurement of the dynamic system in Figure
6. True measurements possess prescribed statistics
related to the watermark signal, while fake measurements
do not. Therefore, by checking if a sequence of
current measurements possesses the prescribed statistics,
one can detect cyberattacks on sensor measurements.
Next, we illustrate how to check the prescribed
statistics in measurements.
For simplicity, we consider the linear time-invariant
(LTI) system with single-input, single-output (SISO) illustrated
in Figure 7, which can characterize a broad class of
dynamical systems. The input of the LTI system is the
TABLE 1. Vulnerable measurements in control
schemes of a future distribution system.
Control Schemes Vulnerable Measurements
Grid-following/gridforming
control
Three-phase instantaneous terminal
voltage and current measurements
(vabc and iabc in Figures 3 and 4)
Secondary control Bus voltage magnitude frequencies
at some critical buses of a microgrid
(E and ~ in Figure 5)
Tertiary control
Real and reactive power exchange
between a microgrid and its host
distribution system (P and Q in Figure 5)
Command
+
Watermark
Watermark
+
Controller
Dynamic
System
Measurement
Figure 6. The general concept of the dynamic watermarking technique.
IEEE Electrification Magazine / SEPTEMBER 2021
59
IEEE Electrification - September 2021

Table of Contents for the Digital Edition of IEEE Electrification - September 2021
