Securing the Internet of Things from the Bottom Up Using Physical Unclonable Functions

Abstract

Cyberattacks that target hardware are becoming increasingly prevalent. These include probing attacks that aim at physically extracting sensitive information including cryptographic keys from non-volatile memory. Internet of Things devices that communicate with the Cloud are susceptible to such attacks. Therefore, the integrity of data and ability to authenticate devices are threatened. Physical Unclonable Functions (PUFs) offer a countermeasure to such attacks. A market analysis of products containing PUFs was carried out. An extract of the market analysis and the inferences that were drawn from it is provided. The analysis showed that although many different types of PUFs have been integrated into a variety of devices, most of them are still used in very rudimentary ways.

Full text

Securing the Internet of Things from the Bottom Up
Using Physical Unclonable Functions
Leah Lathrop∗, Simon Liebl∗, Ulrich Raithel†, Matthias S¨
ollner∗and Andreas Aßmuth∗
∗Technical University of Applied Sciences OTH Amberg-Weiden, Amberg, Germany,
Email: {l.lathrop |s.liebl |m.soellner |a.assmuth}@oth-aw.de
†SIPOS Aktorik GmbH, Altdorf, Germany, Email: ulrich.raithel@sipos.de
Abstract—Cyberattacks that target hardware are becoming in-
creasingly prevalent. These include probing attacks that aim
at physically extracting sensitive information including cryp-
tographic keys from non-volatile memory. Internet of Things
devices that communicate with the Cloud are susceptible to such
attacks. Therefore, the integrity of data and ability to authenticate
devices are threatened. Physical Unclonable Functions (PUFs)
offer a countermeasure to such attacks. A market analysis of
products containing PUFs was carried out. An extract of the
market analysis and the inferences that were drawn from it is
provided. The analysis showed that although many different types
of PUFs have been integrated into a variety of devices, most of
them are still used in very rudimentary ways.
Keywords–Cloud; Physical Unclonable Function; Critical In-
frastructure; Internet of Things; Hardware Security.
I. INTRODUCTION
The Internet of Things (IoT) has engulfed many aspects
of industrial sectors and the lives of private individuals. The
number of actively connected IoT devices is forecast to grow to
22 billion by 2025 [1]. The Industrial IoT (IIoT) is the subset of
the IoT that is used in industrial applications, e.g., healthcare,
energy supply, transportation, and manufacturing. IIoT devices
provide many advantages for traditional systems including
making their management more efficient. The number of IIoT
devices has risen from 3.96 billion in 2018 to 5.81 billion in
2020 [2].
Many IoT devices are constrained by power consumption
and computational resources. The role of IoT devices can
be leveraged through a symbiotic relationship with cloud
computing to carry out data storage, analysis, and monitoring.
In healthcare, storage and analysis of patient data collected by
IIoT devices in the Cloud can be used to avoid preventable
deaths, e.g., by hospital error; real-time monitoring enables
emergency response when necessary [3]. Cloud computing is
also used for the identification and authentication of actuators
according to Molle [4]. The actuators are IIoT devices that
can, e.g., be used to open and close valves to control the water
supply.
The number of opportunities for cyberattacks grows with
the number of IoT devices. The integrity of the data the Cloud
and the IIoT device receive from each other is contingent upon
the security of these devices. The examples in the previous
paragraph showed that IIoT devices are even being used in
healthcare and water supply, which are considered critical
infrastructures in most countries. Compromise or failure of
these systems could harm a society. Attacks that target both
hardware and software threaten these devices. Hardware se-
curity is becoming increasingly important. An example of a
hardware attack is the probing attack, which aims to extract
sensitive data from a device’s non-volatile memory. Physical
Unclonable Functions (PUFs) are a countermeasure to these
attacks.
The motivations for the use of PUFs are elaborated in
Section II. A detailed explanation of PUF technology is given
in Section III. An explanation of the applications for which
PUFs can be used in IIoT devices specifically are provided in
Section IV. An extract of a market analysis, which was carried
out on PUFs, is presented in Section V. The paper is concluded
in Section VI.
II. MOT IVATIO N
Probing attacks can be used to extract sensitive information
including cryptographic keys from non-volatile memory. The
casing of an Integrated Circuit (IC) is removed, and the internal
wires of a security critical module are accessed to retrieve
the data. A Focused Ion Beam (FIB) uses ions at high beam
currents to remove or deposit chip material with nanometer
resolution. The attacker can use a FIB to deposit conducting
paths that may serve as electrical probe contacts [5]. Tarnovsky
carried out an attack to probe the firmware of the Infineon
SLE 66CX680P/PE security/smart chip by probing the buses
of the chip using an FIB [5] [6]. An informative introduction
on probing attacks can be found in chapter 10 of a book on
hardware security by Bhunia and Tehranipoor [5].
Hardware attacks, such as probing attacks, may need more
knowledge, time, and monetary resources than software related
attacks. However, they must still be considered a valid threat.
The attacks are more accessible than some may assume. An
FIB can be purchased on the resale market relatively inex-
pensively or rented at an hourly rate. Furthermore, there are
parties for which the above stated factors are not a hindrance.
Politically motivated attacks including cyberwarfare must be
taken into consideration when evaluating the security of IIoT
devices employed in critical infrastructures. Such attacks have
occured in the past and may be state-sponsored, eliminating
time and financial resources as obstacles. Examples of attacks
on critical infrastructures include two attacks that resulted in
power outages in the Ukraine. In December 2015, a cyberattack
on three Ukrainian energy companies rendered approximately
225,000 people without power for several hours [7]. Ukraine’s
top law enforcement claimed this was a cyberattack by Russia.
Investigations following the attack showed evidence to support
the claim [8]. A second attack took place almost exactly a year
later [9]. The attacks on the power supply in the Ukraine were
44Copyright (c) The Government of Germany, 2020. Used by permission to IARIA. ISBN: 978-1-61208-778-8
CLOUD COMPUTING 2020 : The Eleventh International Conference on Cloud Computing, GRIDs, and Virtualization

not caused by hardware attacks on IIoT devices. However, IIoT
devices are employed to take on various roles in energy supply.
If attackers retrieve a cryptographic key from such a device,
they may be able to eavesdrop onto the communication with
the Cloud. This can help them gain information that will aid
them in an attack.
A malevolent faction may go about an incursion on crit-
ical infrastructures with so much exertion because of the
considerable amount of damage that can be caused. Denial
of Service (DoS) attacks on power supply, which is usually
also considered a critical infrastructure, can have detrimental
effects on the economy. The authors of [10] created blackout-
simulator.com, a tool that provides an estimate of the economic
damage caused by power outages in Europe. The user can
specify the start time, date, and the length of a power outage,
and the region in which the power outage is taking place and
receives an estimate of the economic damage. For example,
the tool estimated the damages caused by a hypothetical six
hour power outage in the state of Bavaria starting at 8 am on
February 24th, 2020 to be approximately 660 million euro [11].
Furthermore, other critical infrastructures, such as healthcare,
would also break down, causing deaths. This provides another
reason why it is important to defend against all cyberattacks
on IIoT devices, especially those that are used in critical
infrastructures.
Some may also consider the shrinking size of integrated
circuits with time a limiting factor. However, FIBs are also
used for the failure analysis in ICs and will therefore continue
to be developed and researched to accommodate the size of
hardware [5].
A recent study shows that hardware- and silicon level
security are becoming a reality for many companies. Forrester
Consulting was commissioned to carry out a study to evaluate
the needs of companies managing breaches to their hardware-
and silicon-level devices and supply chains. The survey was
carried out recently — between March 2019 and May 2019
— and included decision makers in 307 companies. The
study showed that 63% of companies had experienced data
compromise or breach due to an exploited vulnerability in
hardware or silicon level security at least once within the last
12 months; 70% of the surveyed companies consider silicon-
level security as critically important or very important [12].
The broad spectrum of invasive and non-invasive hardware
attacks were implied by this study. These also include probing
attacks.
IIoT devices are particularly susceptible to hardware at-
tacks for several reasons. Man-At-The-End (MATE) attacks
happen from the inside when an adversary gains “physical
access to a device and compromises it by inspecting or
tampering with the hardware itself or the software it contains”
[13]. Several different third parties, some of which are trusted,
have unhampered access to IIoT devices at various points in
their life cycle. Companies have their IC designs manufactured
in semiconductor fabrication plants. There are some cases in
which the manufacturer must place information including cryp-
tographic keys into non-volatile memory during production.
The manufacturers may try to extract the information and
keep it. During operation, (I)IoT devices are often employed
in remote areas without supervision giving attackers unlimited
access to the device.
III. PHYSICAL UNCLONABLE FUNCTIONS
PUFs offer a countermeasure against probing attacks. Anal-
ogous to biometrics, such as fingerprint detection or a retinal
scan, a probabilistic characteristic of a physical object is
used to derive a unique cryptographic secret. Semiconductor
components in electrical devices contain production tolerances,
which are usually unwanted and cannot be controlled. Al-
though these tolerances are only visible on a microscopic
level, they manifest themselves in small differences in physical
sizes, e.g., two voltages may be slightly different. Therefore,
devices which are constructed in exactly the same way can
be individualized. PUFs that derive their fingerprints from
tolerances from the semiconductor production process, e.g.,
random fluctuations in the dopant concentration or doping
profiles, are called silicon PUFs.
A wide variety of different PUFs have emerged including
the arbiter PUF. Figure 1 shows how a single bit can be
derived to illustrate the principle of the arbiter PUF. A chain of
electrical components, each having two inputs and outputs, is
formed resulting in two race tracks for electrical signals. When
applying an electrical signal to both inputs at exactly the same
time, the signals should theoretically arrive synchronously.
Contrary to what might be expected, the arrival times of
the electrical signals are minutely different, due to tolerances
from the semiconductor production process. The outputs are
encoded as a “0” or a “1,” and the bits for the keys can
be derived based on which output the signal arrived at first.
The output of a PUF is called the response [14]. A third
input allows for configuration of the paths; each electrical
component can either be configured as straight or switched.
Different configurations for PUFs are called challenges. Pairs
of challenges and responses are called Challenge-Response
Pairs (CRPs).
Arbiter
...
b0=1 b1=0 bk-2=1 bk-1 =0
Figure 1. Arbiter PUF [14].
The SRAM PUF, first introduced by Guajardo in [15],
offers another method of deriving a cryptographic key from
a stochastic process. SRAM cells are constructed in a way
that enables easy writing making them prone to intrinsic
fluctuations. This does not affect the SRAM cells in any way
during normal operation when an externally exerted signal is
applied to them. However, when the memory cells are in an
undefined state, they take on disparate values. Since SRAM
is a form of volatile memory, such a state is achieved during
power-up. The cryptographic secret can be extracted from the
device during that time. The response is retrieved from the
states of the memory cells of the SRAM.
The SRAM and arbiter PUFs can both be considered
silicon PUFs. Although they share this similarity, these PUF
types can also be distinguished in several different ways. The
nature of the probabilistic behavior is different. PUF principles
that are derived from similar processes can be separated into
different categories. The SRAM PUF belongs to the category
of memory-based PUFs which are derived from a type of
memory. The arbiter PUF belongs to a category of PUFs that
45Copyright (c) The Government of Germany, 2020. Used by permission to IARIA. ISBN: 978-1-61208-778-8
CLOUD COMPUTING 2020 : The Eleventh International Conference on Cloud Computing, GRIDs, and Virtualization

rely on delays of signals called delay-based PUFs. The second
large difference lies in the amount of challenges which can be
applied to a PUF instance. A PUF which has very few or only
one challenge is called a Physically Obfuscated Key (POK) in
PUF literature. A PUF that has many challenges is called a
PUF or strong PUF. The SRAM PUF is an example of a POK
because it only has one challenge whereas the arbiter PUF
has many challenges and can therefore be considered a strong
PUF. The difference is important when considering how they
are integrated into security protocols.
Many formal definitions have been introduced in literature.
The definitions provided by R¨
uhrmair are best suited for IIoT
devices because of the consideration that the adversary has
access to the device for a long time [16]. The source provides
formal definitions for both POKs and strong PUFs. Assuming
an adversary has access to a PUF for a set amount of time and
can retrieve CRPs from it. A PUF is strong if the adversary
can not collect enough CRPs, to deliver the correct response
to a randomly chosen challenge with a probability greater
than 90%. The probability must be greater than 50% to allow
systems with binary outputs to be strong PUFs. However,
whether that value is 90% or 75% is somewhat arbitrary. The
key derived from a system may be called obfuscating PUF or
POK if it derived at least in part from random, uncontrollable
manufacturing variations. It must also be infeasible for an
attacker to guess each digit in the key with a probability greater
than 90% when given the device for a specified amount of time
[16].
IV. APP LI CATIONS OF PUFS
A PUF key can be used to hide a cryptographic key, thereby
eliminating the risk of a probing attack. PUF keys are not
stored in the device but are generated on demand when they
are needed and subsequently deleted. A cryptographic secret
can be derived from a PUF and used directly to substitute
one which was stored in non-volatile memory. The PUF
response can alternatively be used as a Key Encryption Key
(KEK) to encrypt sensitive information stored in non-volatile
memory including cryptographic keys. In the former scenario,
an attacker would no longer find the cryptographic secret in
the device when it is powered off. A probing attack in the
latter scenario would be futile because the data is encrypted.
A POK is well-suited for this because there is no need to
store a challenge. Security protocols that are not specifically
designed for PUFs can then be used.
Several protocols that harness the specific advantages of-
fered by PUFs have also been designed that offer improve-
ments upon traditional security protocols. These include proto-
cols for authentication and authenticated execution for a variety
of different devices on a spectrum of capabilities regarding
power consumption and computing power. A protocol, which
is based on the principle of the Controlled PUF (CPUF), was
introduced by Gassend in [17]. A CPUF can only be accessed
through an algorithm that is physically linked to the PUF in an
inseperable way. The algorithm can be used to restrict chal-
lenges or limit information about responses. The algorithms
with which the PUF can be accessed in this particular protocol
are shown in Figure 2. The owner of the PUF has one CRP that
was extracted from the PUF before it was employed. This CRP
was extracted by applying a pre-challenge to get a response.
The actual challenge can then be computed by calculating the
hash value of the combination of the pre-challenge and a hash
HPUF
H
H
H
PUF
response
secret
program
program
prechallenge
challenge
getresponse(prechallenge)
getsecret(challenge)
Figure 2. Algorithms to access the CPUF [17].
value of the program, as shown in the box with the dotted
line. This allows the owner to calculate the same secret the
PUF calculates using getsecret(prechallenge)later on. In
IoT devices this protocol could, e.g., be used to authenticate a
measurement taken by a sensor. The execution program is sent
to the PUF. The first instruction of the program is to calculate
getsecret(prechallenge). The secret can then be used by
the device to generate a Message Authentication Code (MAC).
The owner of the PUF can also generate the secret because he
has the CRP and can use it to verify the result. Even if attackers
extract the challenge from the program, they will still not be
able to use it because they need the pre-challenge to calculate
the response. This is impossible because it would require them
to reverse the hash functions.
Although this paper focuses on the use of PUFs to secure
(I)IoT devices communicating with a cloud, there are also
many other cloud applications in which PUFs can be used.
A Field-Programmable Gate Array (FPGA) can be used to
accommodate hardware to accelerate certain algorithms, e.g.,
in cryptography. They can be reprogrammed offering flexi-
bility. There are services that offer their customers to carry
out their work on FPGA boards in the Cloud. These services
include Amazon Web Service’s EC2 F1 Instances and services
offered by the company reconfigure.io. The CPUF could be
used to authenticate the results of the computations of the
FPGA boards.
V. AVAILABLE PUF TECHNOLOGIES
Several PUF technologies have emerged on the market. An
extensive analysis of available PUF technologies was carried
out. An extract of the results of the market analysis is provided
below. The PUF technologies that were included in the extract
were chosen because they best illustrate the insights gained
in the market analysis. Most companies that integrate a PUF
into their products buy the technology from a vendor as
Intellectual Property (IP). The IP vendors were researched
and mapped to the companies that integrate them into their
products. In this way, a better idea could be gained of all
available PUF technologies because not all IPs have been
integrated into products that are available for public purchase.
An insight could also be gained into what companies license
their technologies from the same vendor.
When contemplating the integration of technologies with
PUFs into IIoT devices, there are several challenges that
46Copyright (c) The Government of Germany, 2020. Used by permission to IARIA. ISBN: 978-1-61208-778-8
CLOUD COMPUTING 2020 : The Eleventh International Conference on Cloud Computing, GRIDs, and Virtualization

must be considered. For example, IIoT devices can sometimes
have longer lifespans in comparison to regular IoT devices.
Therefore, it is even more important that these can be flexibly
updated because it is far more likely that changes in security
may occur over a span of ten years than over a span of two
to three years.
The IP vendor Intrinsic ID designs a PUF IP that uses the
SRAM PUF technology as described in Section III. The tech-
nology can either be integrated into a product as a hardware IP
(QuiddiKey) [18] or software IP (BroadKey) [19]. BroadKey
can even be integrated into devices that have already been
employed such as IoT devices [20]. The company Renesas
has a family of Microcontroller Units (MCUs) called Synergy.
Renesas offers a free version of BroadKey called DemoKey
which can be tested on Synergy MCUs [21]. Several vendors
of electrical components have integrated the hardware IP
QuiddiKey into their products. These include NXP’s LPC5500
series of MCUs [22] and the LPC540XX family of MCUs
[23]. NXP also includes the PUF in two families of i.MX RT
crossover processors — the i.MX RT600 [24] and the i.MX
RT1170 [25]. Crossover processors combine the advantages of
high end MCUs and application processors to meet the needs
of IoT devices [26]. The NXP products use the PUF to encrypt
data in memory and as a KEK to secure cryptographic keys in
non-volatile memory [24] [27]. Microsemi also uses the SRAM
PUF technology in several products including the PolarFire
FPGA Boards to secure non-volatile memory [28].
Two different PUF technologies are based on the principle
of the current mirror circuit shown in Figure 3, the current
mirror PUF by Invia and ChipDNA by Maxim Integrated.
The black portion of the circuit shows a current mirror as it
can be found in many electrical circuits as a constant current
source. The gate and the drain of MOSFET M1are connected.
Therefore, the MOSFET stays in saturation and the current I1
will stay constant.
The gates of M1and M2are connected causing their
potentials to be equal. Equation (1) can be used to calculate
the drain current of a MOSFET [29]. Wand Lare the width
and length of the channel of the MOSFET, VTh is the threshold
voltage, VGS is the gate-source voltage, Cox is the gate oxide
capacitance per unit area, and µnis the charge carrier effective
mobility. If the MOSFETs that are used for M1and M2are
of the same type and from the same manufacturer, the values
of these variables should theoretically be the same. Therefore,
Iref and I1should also be the same. In practice, there will be
small tolerances from the production process, that can affect
any of the variables in (1) and cause miniscule differences
between the two currents. The blue part of the circuit shows,
that a second constant current source can simply be added by
including another MOSFET M3. Small production tolerances
in the MOSFETs will also affect the currents I1and I2.
ID=1
2µnCox
W
L(VGS −VTh)2(1)
The company Invia has developed a PUF as a hardware
IP that utilizes the principle of the current mirror. The PUF
consists of a matrix of cells that each contain two MOSFETs
producing two constant current sources. The matrix consists of
128 elements — 8 rows and 16 columns. Figure 4 shows how
the value of each element is evaluated; only the first row of the
matrix is depicted. The two resulting currents are compared.
The result will depend on which current is larger. There are
Vcc
M1M2M3
Vcc
R1
Iref I1I2
R2
Vcc
Figure 3. Current mirror as a constant current source.
Vcc
Iref
...
...
...
I1I2
Selector
PUF Element (1,1) PUF Element (1,16)
Comparator
I2>I1(Bit = 1)
I2≤I1(Bit = 0)
Figure 4. Current mirror PUF by Invia [30].
currently no products available for purchase to the public that
contain this hardware IP.
Maxim Integrated is a vendor of electrical components.
Rather than purchasing their PUF as a hardware IP, they
have designed their own technology called ChipDNA, shown
in Figure 5. The PUF also makes use of the principle of
the current mirror. It consists of a matrix of 256 elements
— 16 rows and 16 columns. As Figure 5 shows, each cell
contains two MOSFETs, a p-channel MOSFET M1and an
n-channel MOSFET M2. The left part of the circuit and
M1of each element of the matrix form a current mirror,
providing a constant current source. The gate and the drain
of MOSFET M2are connected causing the MOSFET to stay
in the saturation region and switched on. When a MOSFET is
switched on, it conducts current but has a resistance RDS,on so
there is a voltage drop across the component. These voltages
will vary slightly depending on production tolerances of the
MOSFETs used for the current mirror and for the voltage drop.
Therefore, they can be compared in order to derive a value.
In a diagram of the PUF provided by Maxim Integrated, the
gate and drain of Mref are not connected [31]. The assumption
is made that this is a mistake because the circuit would cease
to function if this would not be the case. The 256 elements
of the array are combined into 128 pairs to achieve higher
stability [31]. Maxim Integrated is the assignee of a patent
that describes an algorithm in which matrix elements are paired
[32]. It is likely that this algorithm is used to create the pairs
mentioned in [31].
ChipDNA has been integrated into several electrical com-
ponents, including the DS2477 [33] and DS28E50 [34], which
can be used for authentication. Authentication is carried out
using a challenge-response protocol. The shared secrets needed
for the challenge-response protocol can be stored on the
components. Only one secret can be stored on the DS28E50,
47Copyright (c) The Government of Germany, 2020. Used by permission to IARIA. ISBN: 978-1-61208-778-8
CLOUD COMPUTING 2020 : The Eleventh International Conference on Cloud Computing, GRIDs, and Virtualization

M1M1
M2M2
V1V2
... ...
Iref
Mref
PUF Element 1 PUF Element 2
Figure 5. ChipDNA PUF Technology by Maxim Integrated [31].
and multiple secrets can be stored on the DS2477. The key
generated by the ChipDNA is used to “cryptographically
secure” all data stored on the device including the secret for the
challenge-response protocol. The electrical components both
contain a True Random Number Generator (TRNG) and a
SHA-3 engine, which can be used to create a Hash-based
Message Authentication Code (HMAC) for the challenge-
response protocol. Maxim Integrated sees great potential in
the integration of these devices into medical equipment among
other applications. They also offer an MCU with a PUF —
MAX32520 [35]. The PUF can be used for internal flash
encryption, device authentication, and to generate a public and
private key pair. The associated public key can be exported and
signed by a certification authority.
According to several sources, the PUF that is integrated
into Xilinx products is sourced from the IP designer Verayo
[36]. The PUF technology that Xilinx integrates into their Zync
UltraScale+ products is a Ring Oscillator PUF (ROPUF) [37].
It can therefore be deduced that Verayo develops an ROPUF.
Srini Devadas who founded Verayo supervised the masters
thesis in which the ROPUF was introduced [38] and was
involved in a publication in which a variation of the ROPUF is
proposed [39]. In the Zync UltraScale+ products, the PUF is
utilized as a KEK to encrypt a user key. The user key can be
used to encrypt the boot image [37] [40, pg. 270]. The Zync
UltraScale+ products include multi processor system on chips
(MPSoC).
Figure 6 shows a diagram of the ROPUF. An asyn-
chronously oscillating loop is formed by inverting the output
of a digital delay line and feeding it back to the input.
The frequency of the oscillator is determined by the delay
line, which is influenced by the manufacturing tolerances of
the electrical components. Consequently, the instances of the
circuit have distinct frequencies. The edges of the signal are
counted using a digital counter to derive a PUF response. The
function n(t)is the edge count as a function of time. The
input challenge can be used to configure the delay line [41].
In [39], a variation of the ROPUF is introduced to reduce
the influence of environmental variations like temperature. The
counters of two instances of the ROPUF circuit are compared
to derive the bit, instead of using the counter as a response
directly. The exact version of the ROPUF that is used in the
Zync UltraScale+ products is not specified in the datasheet. It
is safe to assume that a variation of the ROPUF is used that
does not require a challenge as there is no mention of this in
the data sheet [40].
Several important insights were gathered from the market
analysis. A variety of different PUF technologies (e.g., current
mirror PUF, SRAM PUF, ROPUF) are incorporated into a
diverse group of devices (e.g., FPGA, MCU, MPSoC). PUFs
delay line edge
detector counter
challenge
n(t)
Figure 6. Ring Oscillator PUF [41].
contribute to the security of a diversity of applications, e.g.,
flash encryption and secure boot processes. However, most are
used in a similar way: to encrypt data on the device or replace
a cryptographic key stored in non-volatile memory. Most of
the technologies still use the PUF in a very rudimentary way,
not taking advantage of the specific PUF properties such as the
CPUF protocol. All of the technologies, which were found in
the analysis, were POKs. Although the ROPUF can potentially
have multiple challenges, no mention of these were made in
the datasheet of the product leading to the assumption that the
ROPUF was implemented without them [40].
VI. CONCLUSION AND FUTURE WORK
Hardware attacks including probing attacks are a surging
problem to which PUFs offer an attainable countermeasure.
Many different PUF technologies have been integrated into
a variety of products on the market. Most PUF technologies
available on the market are only used to secure keys which
are then used in traditional security protocols. Based on all the
sources found in the market analysis, most products currently
available on the market for public purchase do not leverage a
protocol that exploits the specific advantages offered by PUFs
and all used PUF technologies are POKs. It will be interesting
to observe the future developments of the PUF market.
ACK NOW LE DG EM EN T
The research project “Intelligent Security for Electrical
Actuators and Converters in Critical Infrastructures (iSEC)”
is a collaboration of SIPOS Aktorik GmbH, Grass Power
Electronics GmbH and OTH Amberg-Weiden. It is supported
and funded by the Bavarian Ministry of Economic Affairs,
Regional Development and Energy.
REFERENCES
[1] K. L. Lueth, “State of the IoT 2018: Number of IoT devices now
at 7B – Market accelerating,” 2018, URL: https://iot-analytics.com/
state-of- the-iot- update-q1- q2-2018- number-of-iot-devices-now- 7b/
[accessed: 2020-07-27].
[2] “Gartner Says 5.8 Billion Enterprise and Automo-
tive IoT Endpoints Will Be in Use in 2020,” 2019,
URL: https://www.gartner.com/en/newsroom/press-releases/
2019-08- 29-gartner-says-5-8-billion-enterprise-and-automotive-io
[accessed: 2020-07-27].
[3] M. S. Hossain and G. Muhammad, “Cloud-assisted industrial internet
of things (iiot) - enabled framework for health monitoring,” Computer
Networks, vol. 101, pp. 192–202, 2016.
[4] M. Molle, U. Raithel, D. Kraemer, N. Graß, M. S¨
ollner, and A. Aßmuth,
“Security of cloud services with low-performance devices in critical
infrastructures,” in The Tenth International Conference on Cloud Com-
puting, GRIDs, and Virtualization – CLOUD COMPUTING 2019, 2019,
pp. 88–92.
[5] M. Tehranipoor and S. Bhunia, Hardware Security A Hands-On Learn-
ing Approach. Elsevier, 2019.
[6] C. Tarnovsky, “Security Failures in Secure Devices,” 2008,
URL: https://www.blackhat.com/presentations/bh-dc-08/Tarnovsky/
Presentation/bh-dc- 08-tarnovsky.pdf [accessed: 2020-07-27].
48Copyright (c) The Government of Germany, 2020. Used by permission to IARIA. ISBN: 978-1-61208-778-8
CLOUD COMPUTING 2020 : The Eleventh International Conference on Cloud Computing, GRIDs, and Virtualization

[7] “Analysis of the Cyber Attack on the Ukrainian Power Grid,”
2016, URL: https://ics.sans.org/media/E-ISAC SANS Ukraine DUC
5.pdf [accessed: 2020-07-27].
[8] J. Pagliery, “Scary questions in Ukraine energy grid hack,” 2016, URL:
https://money.cnn.com/2016/01/18/technology/ukraine-hack-russia/
[accessed: 2020-07-27].
[9] P. Polityuk, O. Vukmanovic, and S. Jewkes, “Ukraine’s power outage
was a cyber attack: Ukrenergo,” 2017, URL: https://www.reuters.com/
article/us-ukraine- cyber-attack-energy- idUSKBN1521BA [accessed:
2020-07-27].
[10] M. Schmidthaler and J. Reichl, “Assessing the socio-economic effects of
power outages ad hoc,” Computer Science - Research and Development,
vol. 31, pp. 157–161, 03 2016.
[11] “Blackout Simulator 2.0,” URL: http://blackout-simulator.com/ [ac-
cessed: 2020-07-27].
[12] “BIOS Security – The Next Frontier for Endpoint Protection,” 2019,
URL: https://www.dellemc.com/ja-jp/collaterals/unauth/analyst-reports/
solutions/dell-bios- security-the- next-frontier-for- endpoint-protection.
pdf [accessed: 2020-07-27].
[13] M. Jakubowski, P. Falcarin, C. Collberg, and M. Atallah, “Software
protection,” IEEE Software, vol. 28, pp. 24–27, 03 2011.
[14] D. Lim, “Extracting secret keys from integrated circuits,” Master’s
thesis, Massachusetts Institute of Technology, May 2004.
[15] J. Guajardo, S. S. Kumar, G.-J. Schrijen, and P. Tuyls, “Fpga intrinsic
pufs and their use for ip protection,” in Cryptographic Hardware and
Embedded Systems - CHES 2007, P. Paillier and I. Verbauwhede, Eds.
Berlin, Heidelberg: Springer Berlin Heidelberg, 2007, pp. 63–80.
[16] U. R ¨
uhrmair, J. S¨
olter, and F. Sehnke, “On the foundations of physical
unclonable functions,” IACR Cryptology ePrint Archive, June 2009,
2009/277 URL: ia.cr/2009/277 [accessed: 2020-07-29].
[17] B. Gassend, D. Clarke, M. van Dijk, and S. Devadas, “Controlled phys-
ical random functions,” in 18th Annual Computer Security Applications
Conference, 2002. Proceedings., 01 2002, pp. 149–160.
[18] “QuiddiKey,” URL: https://www.intrinsic-id.com/products/quiddikey/
[accessed: 2020-07-27].
[19] “Bk software,” URL: https://www.intrinsic-id.com/products/
bk-software/ [accessed: 2020-07-27].
[20] “Intrinsic ID’s BROADKEY Secures IoT with
Key Management Software Powered by SRAM
PUF,” 2017, URL: https://www.intrinsic-id.com/
intrinsic-ids- broadkey-secures-iot-key-management- software-powered\
\-sram- puf/ [accessed: 2020-07-27].
[21] “Secure Key Management Software,” URL: https://www.
renesas.com/us/en/products/synergy/gallery/partner-projects/
intrinsic-id- secure-key-management-software.html [accessed: 2020-
07-27].
[22] “LPC5500 Series: World’s Arm Cortex-M33 based Microcontroller Se-
ries for Mass Market, Leveraging 40nm Embedded Flash Technology,”
URL: https://www.nxp.com/products/processors-and-microcontrollers/
arm-microcontrollers/general- purpose-mcus/lpc5500- cortex-m33:
LPC5500 SERIES [accessed: 2020-07-27].
[23] “LPC540XX: Power-Efficient Microcontrollers (MCUs) with
Advanced Peripherals Based on Arm Cortex-M4 Core,” URL:
https://www.nxp.com/products/processors-and-microcontrollers/
arm-microcontrollers/general- purpose-mcus/lpc54000- cortex-m4-/
power-efficient-microcontrollers- mcus-with- advanced-peripherals-based\
\-on- arm-cortex-m4-core:LPC540XX [accessed: 2020-07-27].
[24] “RT600 (Datasheet),” URL: https://www.nxp.com/docs/en/data-sheet/
DS-RT600.pdf [accessed: 2020-07-27].
[25] “i.MX RT1170 Crossover MCU Family - First Ghz
MCU with Arm Cortex-M7 and Cortex-M4 Cores,” URL:
https://www.nxp.com/products/processors-and-microcontrollers/
arm-microcontrollers/i.mx- rt-crossover-mcus/i.
mx-rt1170- crossover-mcu-family- first-ghz- mcu-with- arm-cortex-m7\
\-and- cortex-m4-cores:i.MX-RT1170 [accessed: 2020-07-27].
[26] “Crossover Embedded Processors – Bridging the gap between perfor-
mance and usability,” URL: https://www.nxp.com/docs/en/white-paper/
I.MXRT1050WP.pdf [accessed: 2020-07-27].
[27] “LPC55S6x (Datasheet),” URL: https://www.nxp.com/docs/en/
data-sheet/LPC55S6x.pdf [accessed: 2020-07-27].
[28] “UG0753 User Guide PolarFire FPGA Security,” URL:
https://www.microsemi.com/document-portal/doc download/
136534-ug0753- polarfire-fpga- security-user-guide [accessed: 2020-07-
27].
[29] M. T. Thompson, Intuitive Analog Circuit Design. Newnes, 2014,
URL: https://www.elsevier.com/books/intuitive-analog-circuit- design/
thompson/978-0- 12-405866- 8 [accessed: 2020-07-27].
[30] V. Telandro and C. Tremlet, “Why should your next secure design
be PUF based,” 2019, URL: https://www.design-reuse-embedded.
com/ipsocdays/ipsocdays2019/china2019/slides/1-Invia%20-%20Why
should your next PUF based.pptx [accessed: 2020-07-27].
[31] “How ChipDNA Physically Unclonable Function Tech-
nology Protects Embedded Systems (Application Note
6767),” URL: https://pdfserv.maximintegrated.com/en/an/
ChipDNA-Unclonable-Protects- Embedded-Systems.pdf [accessed:
2020-07-27].
[32] P. Parvarandeh and S. Ung Kwak, “Systems and Methods for Stable
Physically Unclonable Functions (US Patent 9,485,094),” 2016.
[33] “DeepCover Secure SHA-3 Coprocessor with ChipDNA PUF
Protection,” URL: https://www.maximintegrated.com/en/products/
embedded-security/secure- authenticators/DS2477.html/tb tab0
[accessed: 2020-07-27].
[34] “DeepCover Secure SHA-3 Authenticator with ChipDNA PUF
Protection,” URL: https://www.maximintegrated.com/en/products/
embedded-security/DS28E50.html [accessed: 2020-07-27].
[35] “ChipDNA Secure Arm Cortex M4 Microcontroller,” URL:
https://datasheets.maximintegrated.com/en/ds/MAX32520.pdf
[accessed: 2020-07-27].
[36] Design&Reuse, “Verayo puf ip on xilinx zynq ultrascale+ mpsoc
devices addresses security demands,” URL: https://www.design-reuse.
com/news/40875/verayo-puf-ip- xilinx-zynq- ultrascale-mpsoc.html?
utm campaign=40875&utm content=1&utm medium=rss&utm
source=designreuse [accessed: 2020-07-27].
[37] E. Peterson, “Developing Tamper-Resistant Designs with
Zynq UltraScale+ Devices (XAPP1323),” 2018, URL:
https://www.xilinx.com/support/documentation/application notes/
xapp1323-zynq- usp-tamper-resistant-designs.pdf [accessed: 2020-07-
27].
[38] B. Gassend, “Physical random functions,” Master’s thesis,
Massechusetts Institute of Technology, February 2003.
[39] G. Suh and S. Devadas, “Physical unclonable functions for device
authentication and secret key generation,” in Proc 44th ACM/IEEE
Design Automation Conference, 07 2007, pp. 9–14.
[40] “Zynq UltraScale+ Device Technical Reference Manual,” 2019,
URL: https://www.xilinx.com/support/documentation/user guides/
ug1085-zynq- ultrascale-trm.pdf [accessed: 2020-07-27].
[41] R. Maes and I. Verbauwhede, “Physically unclonable functions: A study
on the state of the art and future research directions,” in Information
Security and Cryptography. Springer Berlin Heidelberg, 2010, pp. 3–
37, URL: https://doi.org/10.1007%2F978-3-642-14452-3 1 [accessed:
2020-04-10].
49Copyright (c) The Government of Germany, 2020. Used by permission to IARIA. ISBN: 978-1-61208-778-8
CLOUD COMPUTING 2020 : The Eleventh International Conference on Cloud Computing, GRIDs, and Virtualization