Thermal Imaging Attacks on Keypad Security Systems
Wojciech Wodo and Lucjan Hanzlik
Department of Computer Science, Wroclaw Univeristy of Technology, Wybrzeze Wyspianskiego 27, Wroclaw, Poland
Keywords: Thermal Imaging Cameras, Thermal Imaging Attacks, Cash Machine, Cipher, Terminal, Keys, Access
Code, PIN Number, Side Channel.
Abstract: The paper discusses the issue of thermal imaging attacks on a variety of keyboard devices, such as cash
machines, payment terminals, combination locks or computer keyboards. The aim of the research was to
obtain the entered code or password in the most non-invasive way. As it turned out, attacks based on images
from thermal imaging cameras are very easy to carry out and work in almost every case, which calls for
extra safety measures. The authors consider various attack scenarios and come up with recommendations for
both manufacturers and users of electronic keyboard security systems.
1 INTRODUCTION
An observation of development of electronic and
information systems in almost all areas of life makes
one notice a replacement of old solutions based on
mechanical or classic methods of verifying the
authorization or identity. You no longer need a lock
to secure a room, but a digital code; you do not pay
in cash, but use a credit card and authorize the
transaction by entering a PIN number. What is more,
you no longer withdraw money after a long wait in a
line at a bank, but use any of omnipresent cash
machines instead, also validating the withdrawal by
entering your PIN number. It is an understandable
trend resulting from manufacturers’ or contractors’
urge to make things most convenient for their
customers. However, transferring the area of
security to the digital world results in new threats
posed not by thieves who can steal physical
property, but by cybercriminals who use
sophisticated methods to steal not only the property,
but also the identity of an online user.
1.1 State of the Art
Literature mentions various examples of side-
channel attacks which use additional information
from channels such as electromagnetic radiation,
power consumption, or timing attacks to break
security systems. This category of attacks is really
difficult to fight because the designer of one
particular component of a security system does not
always influence other parts of it or the terms of its
use. Owing to this fact, it was possible to carry out
an attack to extract a RSA private key after
unauthorized monitoring of processor computing
time for particular mathematical operations. One of
such weaknesses was described in a paper by Paul
Kocher, the pioneer of side-channel attacks (Kocher,
1996). Other researchers carried out a power
analysis using a microprocessor and achieved a data
leakage from a smart card, which is mentioned in
(Messerges et al., 1999). The issue of a simple
power analysis – SPA, and a differential power
analysis – DPA, has been researched also by the
authors of this paper (Kocher, 1999). Using
emissions of electromagnetic fields to break security
systems was described in the iconic work by Van
Eck (Eck, 1985); a series of papers on
electromagnetic field analysis in the context of
random number generators, smart cards and hard
drives has been published by Cambridge University
authors (Markettos et al., 2009; Li et al., 2005;
Markettos et al., 2004).
In the case of attacks using a thermal imaging
camera, two main publications can be mentioned: a
strictly demonstrative study by Michal Zalewski
(Zalewski, 2005) which describes cracking a
combination lock in a safe, and a scientific work
from the University of San Diego, California
(Meiklejohn and Savage, 2011), in which the authors
discuss the attack of capturing PIN numbers of
customers using cash machines. The first work
shows that attacks based on temperature distribution
458
Wodo, W. and Hanzlik, L.
Thermal Imaging Attacks on Keypad Security Systems.
DOI: 10.5220/0005998404580464
In Proceedings of the 13th International Joint Conference on e-Business and Telecommunications (ICETE 2016) - Volume 4: SECRYPT, pages 458-464
ISBN: 978-989-758-196-0
Copyright
c
2016 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
examination on the surface of a digital lock after
entering a PIN (touched by a human) are possible to
carry out. The authors of the second work analyze
details of the attack, consider various scenarios and
automate the process of PIN recovery from
fragments of recorded thermal imaging footage.
However, they only focus on cash machine
keyboards with plastic keys covered with rubber,
claiming that due to the high conductivity of the
metal used in other models of keyboards it was
impossible to obtain any useful data.
1.2 Motivation
Taking into account the dates of the above
mentioned publications describing thermal imaging
attacks and their selective nature, we decided to
examine what has been done to date by
manufacturers of such sensitive devices as cash
machines, safes or payment terminals in order to
ensure their users’ safety. We also took up a
challenge to analyze attacks on devices made from
metals with high conductivity and a high level of
reflection of thermal radiation excluded from the
study by the mentioned authors. The most important
premise was to design and carry out attack scenarios
in which the victim would be oblivious (no
interaction with the victim), and the attacker would
use stealth (the attacker is always hidden). We
adapted this attitude because we deeply believe that
in the digital era of today such criminals are the
most dangerous and hardest to identify.
1.3 Structure of the Paper and
Contribution
Chapter One constitutes an introduction to the
subject of thermal imaging attacks, describing our
motivation and discussing other works on the issue,
proving that it poses a real threat. In Chapter Two of
the work we present several scenarios of thermal
imaging attacks devised by ourselves which we have
simulated on various security devices, from
computer keyboards and digital door locks to cash
machines and payment terminals. Building on the
observations from the previous chapter, in Chapter
Three we give recommendations for manufacturers
and users of electronic security systems, which
could lower the potential threat of thermal imaging
attacks. The last and binding part is Conclusions
which constitutes a conclusion to our findings and
suggests further areas of investigation with thermal
imaging as a source of information for security
breaches.
Figure 1: A micro camera in a cash machine case.
Source: http://krebsonsecurity.com/.
2 ATTACK SCENARIOS
Undoubtedly, the majority of cash machine users
know the technique of a skimmer’s attack on ATM
cards (Snapsupplies.com, 2014), i.e. copying ATM
cards by installing skimming devices, micro cameras
(Fig.1) or keyboard cover plates (Fig.2) recording
the entered PIN number. These attacks, however,
require an invasive interaction with the cash
machine (a partial disassembly of its case or
installation of additional parts); approaching the
cash machine itself or any non-standard behavior at
it may look suspicious to security guards or potential
bystanders. Careless installation of skimming
devices or cameras can be easily noticed and
neutralized, with the skimmer traced owing to
surveillance cameras footage.
Figure 2: A keyboard cover plate. Source:
http://i.dailymail.co.uk/.
Thermal Imaging Attacks on Keypad Security Systems
459
Figure 3: Attack on a computer keyboard, time: 5 seconds, PIN:8962.
Nowadays, prices and sizes of thermal imaging
cameras are significantly lower than in the past and
the devices no longer cost several thousand dollars.
It is worth mentioning that some manufacturers
produce special modules which are compatible with
smartphones and enable thermal imaging in your
own phone; e.g. Seek Thermal offers micro USB
thermal imaging cameras for as little as ca. 250
USD. For comparison purposes we present our
thermal imaging camera used in the research -- Testo
Thermal Image 890 which works with the frequency
of 33 Hz and a standard 640 x 480 resolution,
expanding to 1280 x 960 in the Super Resolution
mode.
Our idea of the attack is to remain entirely
anonymous for the victim with a simultaneous lack
of modification in the attacked device (an oblivious
attack). The adversary takes the role of a potential
user of a chosen device and carries out a standard
procedure of its use without giving rise to any
suspicion. In the meantime, they record the thermal
image of the keyboard using a smartphone with a
thermal imaging camera. For a bystander it looks as
if the adversary were queuing or checking something
on their phone.
As part of our research, we carried out several
dozen attacks on various available keyboard devices:
cash machines, combination door locks, computer
keyboards and payment terminals. Conclusions
drawn from these attacks are also valid for other
devices with a similar structure or made from similar
materials, such as intercoms or parcel lockers.
2.1 Attack 1: Computer Keyboard
A basic attack which piqued our curiosity was a
computer keyboard attack – an attempt to intercept a
login password or an authentication code. We
carried out an attack on a standard external keyboard
by Dell. The password was entered by a fast-typing
person and can be easily read using thermal imaging
pictures (Fig.3). With a longer waiting time it is
difficult to establish the order of the pressed keys,
but it is still certain which were used to enter the
password. In the case of built-in laptop keyboards,
part of the keys are heated up by internal systems, so
the temperature of some of them is much higher than
the ambient temperature, sometimes even than
human body temperature. In such cases it is
impossible to identify the pressed keys in our attack
scenario.
2.2 Attack 2: Digital Door Lock
We considered two types of combination door locks:
with a protective film covering the keyboard and
without it. The scenario of the attempt was fairly
simple: we waited for an authorized person to
approach the door, enter the code and go inside. At
the same time we recorded thermal imaging footage.
Even after a long break of 30 seconds, it was easy to
decipher the code digits; with less waiting time it
was not problematic even to establish the order of
the digits (Fig.4).
2.3 Attack 3: Cash Machine
We carried out attacks on cash machines in
cooperation with Bank Zachodni WBK, having been
granted access to cash machines located in the most
popular spots visited by bank customers every day.
We considered the following three locations of those
cash machines: in buildings (heated and
soundproof), in semi-open spaces (mudrooms) and
SECRYPT 2016 - International Conference on Security and Cryptography
460
Figure 4: Attack on a combination door lock (no protective film), time: 5 seconds, PIN:1268.
Figure 5: Attack on a cash machine (metal keyboard keys), time: 1 minute.
outdoors. It was crucial to take into consideration a
variety of temperatures and weather conditions
which could influence the heat flow and its retention
on the touched parts of the cash machine. An
additional aspect was the recording of sounds made
by the cash machine while pressing its keys, crucial
while entering a PIN number.
The first thing we examined was the average
time of the user interaction with the cash machine,
assuming a fast cash withdrawal scenario. The
whole activity lasted 45 seconds, and counting from
the moment of entering the PIN -- 25-30 seconds.
When the user wanted a confirmation of the
transaction, it added 10-15 seconds to the activity. It
is worth mentioning that the used models of cash
machines had metal keyboard keys.
All attempts to acquire thermal imaging with
traces of the entered PIN ended up in a fiasco. Even
in the case of holding two fingers simultaneously on
a metal button and the case for ca. 1 minute, the
result was inconclusive (Fig.5). Even though, the
heat on the cash machine case was still visible.
Having figured out that the reason for the fiasco
was very low emissive power of the metal used in
the keys (e.g. aluminum), we designed a
modification of the attack in which the keyboard
was covered with a thin layer of an insulator before
carrying out more attempted attacks. The mentioned
insulator can be any spray creating a thin film on the
sprayed surface, such as hair spray or sunburn spray.
Another problem that we had to face was the fact
that the presence of people generating heat
influenced the measurement of the keys’
temperature due to the effect of reflection of thermal
radiation off the polished surface of the metal keys.
In further tests we utilized an information kiosk
keyboard with polished metal keys just like in the
case of the tested cash machines. In the first place
Thermal Imaging Attacks on Keypad Security Systems
461
Figure 6: Attack on metal keys (orange box– keys covered with hair spray, red box – keys covered with nail polish), t. 5 s.
we tested various insulators, e.g. stretch foil, hair
spray, transparent nail polish (Fig.6). Properties of
the used materials and various layers which were
created have proven that the most effective and
durable solution was the nail polish. It creates a
transparent, impalpable, durable and thick enough
layer which retains heat and counters the reflection
phenomenon. In the case of the other two solutions
the effect was unsatisfactory – the hair spray created
a layer which was too thin, did not retain the heat for
long, and was prone to mechanical damage. The foil
was easy to notice and did not protect the keyboard
against the reflection phenomenon.
The next step was covering the 0-9 keys with a
layer of nail polish and then carrying out an attack
on the user’s PIN number. We had no problem
getting the correct sequence of the keys after a short
time – the effect lingered for ca. one minute; after
this time the sequence was a little blurry and we had
to try out several combinations. At this stage of the
research we noticed an opportunity for an attack by
cooling a metal keyboard and recording the
increased temperature of the keys pressed by the
user. Such attacks can be carried out using
compressed air. The effect does not linger long,
therefore it can only be used in the next customer
scenario. The advantage of this solution is that it
does not leave any visible traces on the attacked
device.
2.4 Attack 4: Payment Terminal
In the case of payment terminals, we focused on a
more difficult device in our view – a battery-
powered mobile payment terminal. As opposed to its
fixed equivalent in which the keyboard is a
peripheral device, the mobile payment terminal has
an integrated structure. It means that the keyboard is
located in the same case as the other electronics, and
the temperature of the device is high owing to the
heat generated by the microprocessor and other
elements (similarly as in the case of the laptop,
consider Attack 1). We decided to carry out two
types of attacks: one utilizing the fact that the user
comes from the outside where the temperature of
their hand is lower than the temperature of the
terminal, and one assuming that the customer has
been in the building for some time and the
temperature of their hand is comparable to the
normal human body temperature. In both cases we
can easily identify the keys pressed on the terminal
(Fig.7). The worst case scenario is the situation
when the temperature of the terminal is similar to the
temperature of the customer’s fingers; then the
temperature of the keys changes and you are unable
to identify the used keys.
3 ANALYSIS OF THE RESULTS
AND RECOMMENDATIONS
The experimental attacks carried out by us clearly
indicate that the most significant factor for security
is the material of the keys used for entering the PIN
code. If you use a material with high thermal
conductivity, e.g. a metal, the heat from the user’s
finger can spread onto the whole button. However,
owing to another property of the material, its
emissive power (Wisniewski, 2012), thermal
imaging does not show any thermal changes on the
key’s surface. All polished metals (especially
aluminum) have a low emissivity factor, which
causes problems in observing their thermal radiation.
In the case of materials of high emissive power, such
as plastics, ever brief contact with human skin
results in heat spreading on the key’s surface. Due to
the fact that the key is surrounded by the air which is
SECRYPT 2016 - International Conference on Security and Cryptography
462
an insulator, heat lingers on it for a longer time.
Depending on the weather conditions and properties
of the used material, this time will vary.
Nevertheless, it is long enough to carry out the
mentioned attacks (in some cases even one minute
long). Furthermore, materials vary in absorption
capacity and reflection or passing of heat radiation
(Wisniewski, 2012). A particularly problematic
phenomenon is reflection of thermal radiation of
other objects off the material’s surface. In such cases
it may be difficult to establish whether the observed
emission is due to the material or the user’s heat
reflection only.
The most effective security measures against
thermal imaging attacks, as we have proven, is to
heat up all the keys used to enter the PIN code. It
can be done by putting a hand on their surface or
stroking the surface of the keyboard several times.
After such actions the thermal image shows heat on
each of the keys, blurring information about the
pressed ones.
Both physical and empirical information clearly
indicate that using materials with low emissive
power to manufacture device keys protects them
from thermal imaging attacks. In cases of such
keyboards we only suggest checking and rubbing the
keyboard in order to remove a possible insulating
layer spread by an adversary.
Another side channel turned out to be the sound
made by the cash machine while pressing its keys, or
the sound of the pressed keys itself. The volume
intensity in both cases enables the queuing person to
hear when the customer presses the keys. It is
particularly dangerous while entering the PIN
number because owing to the analysis of keystroke
latencies you can deduce some information about the
digits making up the code. It also facilitates the
analysis of thermal imaging pictures, especially
when their sequence is not entirely clear. The
simplest example of the importance of audio
surveillance is the use of a double digit in a PIN (e.g.
1449); in such cases, the recorded keystroke latency
periods will be: long, short, and long again. The
attacker immediately knows that the middle digit is a
double one. We therefore suggest choosing PIN
numbers without double digits and entering the code
with intentional delay (Hanzlik and Wodo, 2013).
The case of customers of cash machines and
payment terminals is particularly important to us
because while in the first of the mentioned devices
metal keyboards are quite common, they are highly
unlikely to be used in payment terminals. If one of
such key elements of payment systems is so prone to
attacks based on PIN code extraction, you need to
take other safety measures. First and foremost, you
have to make provision against unauthorized attacks
which use the victim’s bank card to make a PayPass
payment (utilizing NFC chips) (Hancke, 2005). Such
transactions are made with no necessary
authorization for limited amounts of money, yet in
the case of obtaining the victim’s PIN number
(which is not so difficult, as we have proven) the
safety measure is no longer valid. The only
effective protection is preventing the card from
starting the protocol by placing it in a special
protective case – a Faraday cage. Owing to this
step, no electromagnetic field can influence the card
and initiate communication with it without our
consent. What is more, it might be a good idea to
dedicate a contactless card to a bank account with
significant limits set for transfers and transactions.
4 CONCLUSIONS
As can be seen, thermal imaging attacks still pose a
threat to users’ safety. They can prove particularly
Figure 7: Attack on metal keys (orange box– keys covered with hair spray, red box – keys covered with nail polish), t. 5 s.
Thermal Imaging Attacks on Keypad Security Systems
463
dangerous for people using contactless payment.
In the past, the PIN number itself was not enough
as the card was necessary, too. Now, as we have
proven earlier, owing to the PayPass contactless
payment, it is enough to stand close to the victim
and use their card to carry out the transaction. When
you add the PIN code obtained in a thermal imaging
attack to the equation, the limit set for the amount of
money spent disappears completely (since it is
authorized by the stolen PIN code). In order to avoid
modern cybercriminals, you have to be extremely
cautious and sensible while using technological
devices. It is not enough to protect yourself against
one threat only; you need to take safety measures
against all forms of attacks, and this step requires
knowledge and proactive behavior. The key action
taken by manufacturers and providers of such
modern devices should be raising users’ awareness
of potential threats and how to deal with them.
ACKNOWLEDGEMENTS
We would like to thank Mr. Janusz Rogula, PhD.
Eng. of Institute of Power Engineering and Fluid
Mechanics of Wroclaw University of Technology
for lending measuring instruments necessary to carry
out our research. We would also like to extend our
gratitude to the authorities of Bank Zachodni WBK,
in particular to Mrs. Monika Bejma and Mr. Rafal
Wachowski, for their trust and sharing their cash
machines as research tools. This research was
supported by National Research Center grant
PRELUDIUM no 2014/15/N/ST6/04375.
REFERENCES
Kocher, Paul, (1996). Timing Attacks on Implementa-
tions of Diffie-Hellman. RSA, DSS, and Other
Systems, Advances in Cryptology—CRYPTO’96.
Messerges, Thomas S., Ezzy A. Dabbish, and Robert H.
Sloan, (1999). Investigations of power analysis attacks
on smartcards. In Proceedings of the USENIX
Workshop on Smartcard Technology on USENIX
Workshop on Smartcard Technology (WOST'99).
Kocher, Paul, J. Jaffe, and B. Jun. (1999). Differential
power analysis. Proceedings of CRYPTO'99, vol.
1666, pp. 388-397.
Van Eck, W. (1985). Electromagnetic radiation from video
display units: An eavesdropping risk? Computers &
Security 4, 4.
Markettos, A. T. and S. W. Moore. (2009). The Frequency
Injection Attack on Ring-Oscillator-Based True
Random Number Generators. In Proceedings of
Cryptographic Hardware and Embedded Systems
(CHES). Lecture Notes in Computer Science 5747,
Springer, pp. 317-331.
Li, H., A. T. Markettos and S. W. Moore. (2005). A
Security Evaluation Methodology for Smart Cards
Against Electromagnetic Analysis. In Proceedings of
the 39th IEEE International Carnahan Conference on
Security Technology (ICCST 2005), Las Palmas de
Gran Canaria, Spain, pp. 208-211.
Markettos, A. T. and S. W. Moore. (2004).
Electromagnetic Analysis of Synchronous and
Asynchronous Circuits using Hard Disc Heads. 16th
UK Asynchronous Forum, Manchester, UK.
Zalewski, Mariusz. (2005). Cracking safes with thermal
imaging, [online] Available at: http://lcamtuf.
coredump.cx/tsafe. [Accessed 23 May 2016].
Meiklejohn, Sarah and Stefan Savage. (2011). Heat of the
Moment: Characterizing the Efficacy of Thermal
Camera-Based Attacks. Proceedings of Workshop on
Offensive Technologies (WOOT).
Snapsupplies.com. (2014). ATM Skimming Devices,
[online] Available at: https://www.snapsupplies.com/
Snap/Industry-News/ATM-Skimming-Devices_12.asp
x. [Accessed 23 May 2016].
Stefan Wisniewski, Tomasz S. Wisniewski. (2012).
Wymiana ciepla, WNT.
Hanzlik, L., W. Wodo. (2013). Identity security in
biometric systems based on keystroking. International
Conference on Security and Cryptography
(SECRYPT), Reykjavik, Iceland.
Hancke, Gerhard P., (2005). A practical relay attack on
ISO 14443 proximity cards. Technical report,
University of Cambridge Computer Laboratory.
SECRYPT 2016 - International Conference on Security and Cryptography
464