BIOMETRIC BASED SMART CARD FOR SECURITY
Chunlei Yang
Advanced Technology Department, Ingenico Group, 9 Quai De Dion Bouton -92816 Puteaux, France
Guiyun Tian, Steve Ward
School of Computing & Engineering, University of Huddersfield, UK
Keywords: Biometrics, Security, Smart card, Integration.
Abstract: Fingerprint has been increasingly used in authentication applications. Smart card is becoming more and
more common and is moving toward a multi-function era. The integration of biometric and smart card is a
trend for the future of smart card. As a part of our research project which concerns a novel security card, we
propose to integrate the fingerprint sensor with the smart card instead of the normal solution where the
sensor is installed with a terminal machine. This solution has some advantages regarding security, user
privacy as well as flexibility. In this paper, we study the biometric security and outline our solution. In
addition, in the system authentication decision part, a novel adaptive decision algorithm which combined
with biometrics, PIN (personal identify number) is introduced. This algorithm can be a better trade-off
between user convenience and security.
1 INTRODUCTION
Biometrics refers to the automatic identification or
verification of living persons using their enduring
physical or behavioural characteristics. Biometric
personal authentication uses data taken from
measurements of a person’s body, such as
fingerprints, faces, irises, retinal patterns, palm
prints, voice, signature, DNA, and so on. Since
biometrics has features of “not be lost or forgotten,
unique”, it is increasingly used in security or privacy
needed devices.
As shown in Table 1, different biometric
technology has its merits and weaknesses (Rodrigo
et al., 2003). For instance, retina scanning requires
that a laser is shone onto the back of the eyes and the
unique characteristics of the retina are measured.
The retina is an extremely stable biometrics because
it is ‘hidden’ and not subject to wear, the
system is
hard to fool because the retina is not visible and
cannot be faked easily. However, it is a potential risk
to health and the invasive nature is unattractive to
customers. Face recognition is a quite natural
method, but in practice, it is affected by lighting,
pose and expression strongly. It also needs high
computation power and the embedded system cannot
meet this requirement. Therefore, thinking
comprehensively based on the factors of accuracy,
cost, convenience and marketing, fingerprint has the
feature of convenient, proven, miniaturization and
inexpensiveness, and it has the best potential for
mass market authentication schema.
As in a typical biometrics-based personal
authentication, fingerprint authentication uses a
four-step process including capture, extraction,
Table 1: Comparison of common biometric.
Type Merits Weakness
Iris High accuracy,
hard to fool
Large and expensive
equipment
Face Non-invasive, no
physical interaction
with sensor needed
Low accurateness,
affected by lighting
& face position
Finger-
print
Convenient, well-
developed,
inexpensive, high
potential for
miniaturization
Accuracy depends
on fingerprint
quality,
Finger subject to
wear
Voice Non-invasive and
natural
Subject to wide
variation, hard to
detect recorded
voice
Retina Stable, hard to fool Invasive, not well
tested, expensive
240
Yang C., Tian G. and Ward S. (2005).
BIOMETRIC BASED SMART CARD FOR SECURITY.
In Proceedings of the Second International Conference on e-Business and Telecommunication Networks, pages 240-246
DOI: 10.5220/0001407302400246
Copyright
c
SciTePress
comparison and matching. The pre-stored minutiae
for matching during an enrolment is also called
template. Two techniques are used to decide if the
verification data really corresponds with the
reference data. One is based on minutia matching
(local details) and the other is based on pattern
matching (global structure). Generally speaking,
minutia matching is more commonly used. Figure 1
illustrates how to extract fingerprint minutiae.
Figure 1: Fingerprint minutiae extraction.
There are two common ways to implement a
biometric system according to the different places of
storing templates and matching: online and offline.
Online means the fingerprint templates are stored
and matched in a centralized server computer. This
solution has advantages in terms of management and
rapid system update, however a stable
communication is always needed and it will increase
the cost and slowdown the transaction. Offline
means the authentication can be done locally
because the template is stored and matching is
finished locally. This solution can verify identity
without complex communication infrastructures and
cut cost. It is especially important in mobile
application and at sites away from the
communication line. The vital question for offline
solution is how to store the template securely. A
smart card can be an ideal solution to address these
questions. It can operate both online and offline.
A smart card has an embedded processor and
memory. From a functional standpoint a smart card
is a miniature computer. The smart card has the
capability to record and modify information in its
own non-volatile memory and the security data can
be well protected or ‘hidden’ by the operating
system and hardware. These features make the smart
card a powerful and practical tool against
unauthorized data access and copy (Peyret P. et al.,
1990;
David M. and Moti Y., 2001.). More and more
technologies are integrated with the smart card. The
PKI (public key infrastructure) has reinforced the
smart card security and makes the smart card be an
ideal place to carry varying degrees of sensitive
information. In the past years, the biometric and
smart card technology have been combining together
in some applications (BioT1, 2003; Chunhsing L.
and Yiyi L, 2004). As illustrated in Fig.2, a terminal
with a fingerprint sensor captures the fingerprint and
extracts the minutiae, then the extracted minutiae are
sent to the smart card to match with the stored
fingerprint templates in the smart card. The process
is called match-on-card (MOC) and the card is
called biometric card (BioT3, 2003).
The rest of the paper is organized as follows:
Section 2 analyses the general security of the
fingerprint authentication system, namely attacks
and countermeasures. Section 3 describes the
proposed system, including architecture, procedure
and an adaptive decision algorithm. Section 4 is a
conclusion and future work description.
Fingerprint
capture
Extracted
fingerprint
data
match
decision
Reader (terminal) Smart card
Stored templetes
Figure 2: Diagram of Match-on-Card process.
2 BIOMETRIC SYSTEM
SECURITY
2.1 Fingerprint System Security
In this section, the general security of biometric
system will be analysed.
A generic biometric data processing model is
shown in Figure 3. Within this model, following the
data process from sensor until application, we
identify nine basic biometric attacks (Attack 1; . . . ;
9) that plague biometric based authentication
systems. For simplicity, the enrolment of the
fingerprint template is not included, although that is
a quite important link of the whole biometrics
security system.
Typically, Attack 1 could be an impersonation
attack where the attacker uses a fake fingerprint to
fool the sensor. Attack 2, 4, 7, 8 belong to channel
attacks where the attacker can use line taping,
intercept the biometric data or use previous recorded
signal to replay attacks. Besides such direct channel
attacks, some advanced crypt-analytical techniques,
so called side channel attacks, also pose serious
threats to biometric system even to the channels that
are encrypted. For instance, by analyzing the power
dissipation or timing of encryptions in device,
encrypted information inside can be deduced (An Y.
and David S., 2004.; Ross A. and Markus K., 1997.).
BIOMETRIC BASED SMART CARD FOR SECURITY
241
Attack 3, 5, 6, 9 fall into the categories which attack
the inside software or secure keys (if the
cryptographic technology is employed for secure
data transmission). Below more details about attacks
and countermeasures will be examined.
A fake finger attack is a serious threat to
biometric authentication systems, since this type of
attack directly exploits the intrinsic weakness of
biometrics: easy to be captured and hard to revoke.
When fingers touch an object, the chemicals in
finger sweat may be absorbed into that object (paper
Matsumoto T. 2002 is a good example), and there
are new chemicals which can develop these quite
nicely. Afterwards, a fake finger can be made to fool
the biometric system. With the ongoing development
of technology, a latent fingerprint can be detected
and captured easily and a very sophisticated fake
finger can be made. E.g. the fake print made from
gelatine, which is low-cost, electrically quite like
real flesh, can already fool many optical, capacitive,
pressure based sensors (Matsumoto et al., 2002).
Theoretically each data transfer channel is
susceptible to channel and side channel attacks if it
is not well protected. The typical attacks can be a
replay attack, resubmission of an old digitally stored
biometric signal, or an electronic impersonation.
More specifically, like in Attack 2, after the features
have been captured by the sensor, if the sensor and
the extractor hardware has a long and exposed
channel (e.g. connected with cables), this captured
data can be replaced with a different synthesized
feature set. In Attack 4 the minutiae can be replaced.
In Attack 7 the templates from the stored database
which are sent to the matcher can be altered before
they reach the matcher. In Attack 8, the final
decision of the matching module can be overridden.
From a software perspective, the compiled
source code stored in the system is susceptible to de-
compilation and reverse engineering (Gleb N. and
Nasir M., 2003.), which means the program could be
read and analyzed. Therefore, if the security
mechanism is merely based on some tricks in the
program, it will be easily subverted by analyzing the
program and designing some actions to avoid
triggering the security mechanism. If the adversary
can install a Trojan horse into the biometric system,
some information will be disclosed to the attacker,
etc.
2.2 Countermeasures for Biometrics
Attacks
Based on above threat analysis, some
countermeasures can be taken to improve the
security.
To prevent a fake finger attack, a multi-modal
sensor could be an effective way. In a multi-modal
sensor, for example, in addition to capturing a
fingerprint, the warmth and pulse can also be
detected. Or like some advanced sensor, instead of
taking a static picture of the surface of the finger, it
reads the fingerprint from the live layer below the
surface of the skin. This method ensures that the
device will acquire the fingerprint despite varying
skin moisture levels; abrasion of the fingerprint from
harsh chemicals or friction like rubbing; and
common contaminants such as lotion, grease, or
smoke. This subsurface-imaging approach thereby
eliminates the surface-based recognition failures
common with surface-imaging fingerprint sensors
based on capacitive, thermal, optical, or pressure-
sensing techniques (AuthenTec, 2004).
There are several solutions that can improve the
system security. As proposed in the paper by Nalini
K. et al. (Nalini et al., 2003), 1) “Image based
challenge/response method”. The matcher unit
generates a pseudorandom challenge for the
transaction and the sensor unit acquires a signal at
this point of time and computes a response to the
challenge based on the new biometric signal. 2).
WSQ (Wavelet Scalar Quantization) -based data
sensor
Template
extractor
matcher
Application
Template
database
1 2 3
4
5
6
7
8
9
Legend
x
Data flow
Attacks
Figure 3: Biometric system security model.
ICETE 2005 - SECURITY AND RELIABILITY IN INFORMATION SYSTEMS AND NETWORKS
242
hiding. It uses data hiding techniques to embed
additional information directly in compressed
fingerprint images to guard against replay attacks.
However, such measures can hardly meet high
security requirements. If the hardware is not secured
and the program can be reverse engineering and
analyzed, such a system can be subverted without
difficulty.
Therefore, finally, the essential protection is to
seal as many of the system components as possible
into a tamper-proof device, including the data
transmission channels. If some channels cannot
really be sealed, then cryptographic technology
should be employed to ensure data integrity and
confidentiality. The security key must be very well
protected. Following these thoughts, we consider the
combination of biometrics and smart card could be
an attractive solution. As the match-on-card solution
which is introduced before, the smart card is used to
store the biometric template to match the signals
from outside of the card.
3 PROPOSED ‘CAPTURE &
MATCH-ON-CARD’ SOLUTION
However, we propose another solution which is
other than the above outlined match-on-card
solution. The fingerprint sensor is integrated with
the smart card body in our new proposed system.
The considerations are listed as follows:
- Increase the difficulty for attackers. In practice,
most of the attackers need to install an electric
bug or apparatus to the attacked object. A
terminal machine (card reader) normally has a
spacious plastic housing which contains many
PCBs, electric components, etc. The wires linking
the system components to each other could
become potentially passive or active penetration
routes. It is not difficult to find a small space in
the terminal for installing an electric bug inside.
However, if the fingerprint sensor is integrated
with the smart card, all these electric elements
can be packed into one very thin plastic package,
or even be integrated into one single chip and
interlinks can be hidden.
- Distribute the security risk. The adversary can
get far more potential benefits from
compromising a terminal security system than
compromising a single card. If the biometric
sensor in a terminal is compromised, it will
jeopardize all its users. Thus by distributing the
sensor to the cards can distribute the risk.
- Protect the privacy and increase the flexibility.
Nowadays, the sensor is installed with the
terminal machine. Although the terminal
providers as well as the merchants declare that
“we don’t take your fingerprint images but only
features”, it is hard to believe when the customers
see their fingerprints scanned by the terminal.
Meanwhile, if the card has a biometric sensor
itself, it can improve the flexibility and customers
can use and benefit from the potential advanced
biometric technology everywhere.
Obviously the feasibility of this solution relies on
two issues: the cost of a sensor and whether the
physical shape of the sensor allows it to be
integrated into a card. Thanks to the ever evolving
research and improvements in biometric sensors,
especially the new silicon swipe fingerprint sensor,
this proposed work becomes more realistic than
ever. For example, recently the cost of a swipe
sensor can already be less than four US dollars, and
the dimension and the thickness makes it nearly
possible for it to be integrated into the smart card
body, the thickness of which is 0,8mm. The fragile
sensor can be protected by a thin metal sheet around
it.
The novel security card is illustrated in Fig. 4. In
this paper, we only describe the part of our project
which concerns biometric security. The envisioned
architecture and procedures are presented. In
addition, an adaptive decision algorithm of
authentication, which is combined with biometrics
and PIN (Personal identify number), is introduced.
This algorithm can be a better trade-off between user
convenience and security.
Figure 4: A 3D simulation picture of our project.
BIOMETRIC BASED SMART CARD FOR SECURITY
243
The principle and architecture of proposed
system is illustrated in Fig.5. The smart card has a
fingerprint sensor and a small LED. They are
packaged together and offer the possibility of
integrating the security sensitive components into
one small chip and apply some ripe tamperproof
technologies from the smart card industry
(Wolfgang, 2003). Meanwhile part of the system
security risk can be distributed to many cards.
For our system experiments, a swipe type
fingerprint sensor AES2510 from AuthenTec Inc has
been selected, not only for its small size and low
cost, but also for security. It uses a radio frequency
(RF) imaging technique that allows the sensor to
generate an image of the shape of the live layer of
the skin that is buried beneath the surface of the
finger. Thus it can better prevent attacks like
gelatine fake finger. AuthenTec promised to offer a
smaller and cheaper version of swipe fingerprint
later.
3.1 Architectural Description
As illustrated in Figure 5, theoretically after the
sensor has been integrated with the smart card, all
the capture, feature extraction and matching can be
done inside the card. However, due to the fact that
the normal embedded processor of the smart card as
well as the memory can hardly fulfil the
requirements of complex image processing, the
image data store and fingerprint minutiae extraction
parts are moved to the card reader side. The swipe
fingerprint sensor reads the finger line by line,
generates a challenge and sends the data to FIFO
(first in, first out) via parallel or DMA (direct
memory access) communication. The data in FIFO
will be encrypted and directly sent out to the
memory of the card reader machine. After the image
capture is complete, the image data will be
decrypted and the minutiae extracted before it is sent
back to the smart card for verification. In addition,
one LED light is added and integrated with the smart
card. This LED can change the role of the smart card
from a passive and ‘dumb’ card to an active one,
e.g., it can indicate some serious edicts to improve
the security as well as user convenience.
3.2 Procedure and Security
The authentication procedures are outlined as below
in five steps:
1. Mutual authentication using PKI technology
between the card and the card reader (EMV4.1,
2004)
As shown in Figure 6. The card issuer uses its
Issuer private key S
I
to certify the card public key
P
C,
and saves the certified P
C
in a readable area of
the smart card. However, the card private key S
C
and
the fingerprint template are saved in the ‘hidden’
area in the smart card. These data are hidden, and
cannot be copied or read out by an external card
reader.
The issuer public key P
I
is distributed to the card
reader. So the card reader can use P
I
to verify that
the card’s P
C
was certified by the issuer, and use P
C
to verify the digital signature of the card data.
Therefore, in this way the terminal can confirm that
the card is original and has not been modified. On
the other side, to determine whether the card reader
is genuine, the card can check the certification of the
card reader. In case the above mutual authentication
fails, the application will be cancelled and both the
card reader display and the card will indicate the
error message, i.e. the LED on the card will flash.
This is an important feature because it can detect a
dummy terminal which is made by an adversary to
cheat the user.
FIFO
Minutiae Extraction
Bio template,
private Key S
C
match
Image data
minutiae
Result
Sensor
Challenge
LED
Card reader
Card
Re-process
Encryption/
Decryption
Note:
memory
Certified card
Public key, P
C
Adaptive decision
algorithm
Fi
g
ure 5: Architecture of the
p
ro
p
osed s
y
stem.
ICETE 2005 - SECURITY AND RELIABILITY IN INFORMATION SYSTEMS AND NETWORKS
244
Private Key
(Card)
S
C
Public Key
(Card)
P
C
Private Key
(Issuer)
S
I
Public Key
(Issuer)
P
I
Card reader
P
C
certified
with S
I
S
C
&
fingerpint template
communication
card
Figure 6: Diagram of Dynamic Data Authentication.
2. Session key generation
A session key can be used as a secure key for the
encrypted communication between the card and the
reader (e.g. DES encryption). The session key
derivation function in both the card and the reader,
generate a unique session key Ks for each ICC
application transaction as per the following method.
The system first generates unique Master Keys K
M
from the user primary account number and Issuer
Master key, then Ks can be derived from K
M
, ATC
(Application Transaction Counter) using
diversification data R. The detailed generation
method can refer to EMV definition (EMV 2004).
K
M
:= F (Primary Account Number, Issuer Master
key
)
Ks : = F (K
M,
ATC) [R]
3. Fingerprint capture and extraction
The fingerprint sensor reads the finger image and
adds random data to the fingerprint data. The
random data can prevent replay attack. The mixed
data are sent to FIFO, after DES-encryption using
the session Ks, they are sent out to the memory of
the card reader. After fingerprint reading is
complete, the stored image can be decrypted and the
minutiae extracted. The minutiae are encrypted
again and sent back to the card for authentication.
4. The card decrypts the received minutiae.
5. Match the acquired minutiae with the hidden
fingerprint template in the smart card and generate a
similarity score. The final decision comes from an
adaptive algorithm (refer to section 3.3). The
decision is encrypted and sent both to the card reader
and the smart card LED. This is a special measure
because the conventional way is just to send it either
to the card or the card reader. In this way, even the
attacker faked a result in card reader and the card
reader display shows the operation is right, but the
LED on the smart card will start to flash and give a
warning.
3.3 An Adaptive Decision Algorithm
Two authentication methods, PIN and biometric,
have their own features. The PIN authentication is
stable but prone to be disclosed and forgotten; the
biometric authentication is convenient but cannot
reach a perfect recognition rate and be updated. Thus
they cannot really replace each other completely.
Actually, a high security system can be based on a
combination of three factors: ‘something-you-have’
which is the smart card factor, ‘something-you-
know’ which is the PIN factor and ‘something-you-
are” which is the biometrics factor (Stephen et al.,
2000). Nowadays the smart card becomes a platform
for multi-applications. More and more payment and
non-payment applications (lottery, access control)
have been integrated into a single card. Actually
different applications need different level
authentications. Even the same application, e.g.
payment application, the risk for low amount and
high amount transaction are different.
In order to better balance the security
requirements and user convenience, we propose an
adaptive algorithm and apply it in the authentication
decision. Principally, we first classify different
applications into several predefined levels according
to various security requirements and transaction
value. Then the algorithm selects different methods,
varies the threshold value of biometric similarity
degree, even vary the similarity degree of PIN.
Adaptive decision algorithms are illustrated in
Figure 7.
A new concept of PIN match with a tolerance
(fuzzy PIN) is proposed. For example, for some
applications, when the user can offer a standard
fingerprint, then even he/she makes some small
mistakes in PIN, (e.g. should be 63456 but entered
63455), which the system will also accept (but issue
a warning, so the legitimate user can check it later at
home). To protect the card against exhaustive
search, the card will be locked after 10 successive
unsuccessful fingerprint verifications.
These measures have practical significances and
can cut the management cost. Many calls to the help
desk concern the PIN because it is forgettable.
A lot
of legitimate users’ cards are mis-locked or
applications have to be cancelled on site. A recent
IDC study put annual password management costs at
between US$230-460 per user (BioT2, 2004), which
would add up to a significant amount when a bank
has a large number of customers. The fuzzy PIN
matching combined with biometric measures can
help to avoid such nuisances without lowering the
security.
BIOMETRIC BASED SMART CARD FOR SECURITY
245
low
midium
Inputted
PIN
finger Score>0.85
PIN score>0.95
&
finger score>0.85
PIN score=1
&
finger score>0.95
PIN
match
Fingerprint
match
high
Input
fingerprint
Input PIN &
fingerprint
Input PIN &
fingerprint
Risk rank
PIN warning
Stored
fingerprint
Acquired
fingerprint
Stored
PIN
access
Y
Y
Y
deny
deny
deny
N
N
N
Applications,
amounts
PIN
score
finger
score
4 CONCLUSION AND FUTURE
WORK
This paper has reviewed the security of biometric
system. It argues the advantages of integrating a
fingerprint sensor with smart card, in terms of
security, privacy protection as well as system
flexibility. Based on the review, proposed types of
architecture, procedures and security issues are
presented and analyzed. The merits of the proposed
approach are heightened.
The project is still in progress. Further work on
the system fabrication, implementation and system
evaluation such as system design, minutiae
extraction and testing the system’s real performance,
etc, will be undertaken.
REFERENCES
Anil K. J., Umut U., 2003. Hiding Biometric Data, IEEE
Transaction on Pattern Analysis and Machine
Intelligence, Vol. 25, No. 11, pp. 1494-1498.
An Y. and David S., 2004. Clock-less Implementation of
the AES Resists to Power and Timing Attacks,
International Conference on Information Technology:
Coding and Computing (ITCC'04) Volume 2, Las
Vegas, Nevada, p. 525 A.
AuthTec, 2004. Website of AuthenTec,
www.authTec.com.
BioT1, 2003. Match on card system for IT security,
Biometric Technology Today, Volume 11, Issue 7,
Pages 3-4.
BioT2, 2004. Biometrics secure loan application,
Biometric Technology Today, Volume 12, Issue 7,
Page 3.
BioT3, 2003. Match on card system for IT security,
Biometric Technology Today, Volume 11, Issue 7, pp.
3-4.
Chunhsing L. and Yiyi L., 2004. A flexible biometrics
remote user authentication scheme, Computer
Standards & Interfaces, Volume 27, Issue 1, Page 19-
23.
David M. and Moti Y., 2001. E-commerce applications of
smart cards, Computer Networks, Volume 36, Issue 4,
pp. 453-472.
EMV 4.1, 2004. Integrated Circuit Card Specification for
Payment Systems. Book 2. They can be downloaded
from http://www.emvco.com
Gleb N. and Nasir M., 2003. Preventing Piracy, Reverse
Engineering, and Tampering, COMPUTER, IEEE
Computer Society, Vol. 36No. 7, pp. 64-71.
Matsumoto T. and Matsumoto H. and Yamada K. and
Hoshino S., 2002. Impact of Artificial Gummy Fingers
on Fingerprint Systems, Proceedings of SPIE Vol.
#4677, Optical Security and Counterfeit Deterrence
Techniques IV.
Nalini K. R. and Jonathan H. C. and Ruud M. B., 2003.
Biometrics break-ins and band-aids, Pattern
Recognition Letters 24 (2003) 2105–2113.
Peyret P. and Lisimaque G. and Chua T.Y., 1990. Smart
cards provide very high security and flexibility in
subscribers management, IEEE Transactions on
Consumer Electronics 36 3, pp. 744–752.
Rodrigo de Luis-García, Carlos Alberola-López, Otman
Aghzout and Juan Ruiz-Alzola. 2003. Biometric
identification systems, Signal Processing, Volume 83,
Issue 12, Pages 2539-2557.
Ross Anderson and Markus Kuhn, 1997. Low Cost Attacks
on Tamper Resistant Devices, Proceedings of the 5th
International Workshop on Security Protocols,
Springer-Verlag LNCS No.1361, April 1997, p.125.
Stephen M. and Matyas Jr. and Jeff S., 2000. A Biometric
Standard for Information Management and Security,
Computers & Security, 19 (2000) 428-441.
Wolfgang R., 2003. Overview about attacks on smart
cards, Information Security Technical Report, Volume
8, Issue 1, Pages 67-84 .
Figure 7: Diagram of adaptive decision algorithm.
ICETE 2005 - SECURITY AND RELIABILITY IN INFORMATION SYSTEMS AND NETWORKS
246