TOWARDS BIOMETRIC-BASED AUTHENTICATION FOR
CLOUD COMPUTING
Kok-Seng Wong and Myung-Ho Kim
School of Computer Science and Engineering, Soongsil University, Sangdo-Dong Dongjak-Gu, Seoul, South Korea
Keywords: Biometric-based Authentication, Cloud Authentication System, Privacy Preserving Squared Euclidean
Distance, Data Protection.
Abstract: Cloud computing is an emerging technology that allows different service providers to offer services in an
on-demand environment. Due to the advantages such as flexibility, mobility, and costs saving, the number
of cloud user has increased tremendously. Consequently, a more secure and privacy preserving
authentication system is becoming important to ensure that only the data owner or the authorized user can
gain access and manipulate data stored in the cloud. In the current approach, the service provider
authenticates its users based on the credential submitted such as password, token and digital certificate.
Unfortunately, these credentials can often be stolen, accidentally revealed or hard to remember. In view of
this, we propose a biometric-based authentication protocol, which can be used as the second factor for the
cloud users to send their authentication requests. In our solution, the credential submitted by the users
consists of the biometric feature vector and the verification code. For the user to successful authenticate,
both the biometric feature vector and the verification code must be combined, transformed, and shuffled
correctly. Our proposed solution not only provides the security mechanism for the authentication process,
but also supports the privacy protection for all sensitive information of the user.
1 INTRODUCTION
Cloud computing is an emerging technology which
allows multi-tenant to request for services and
resources from their service providers in an on-
demand environment. It is a complex yet resource
saving infrastructure for today’s modern business
needs, providing the means through which services
are delivered to the end users via Internet access. In
the cloud environment, users can access services
based on their needs without knowing how the
services are delivered and where the service are
hosted.
The US National Institute of Standards and
Technology (NIST) has defined cloud computing as
follows (Mell and Grance, 2009): Cloud computing
is a model for enabling ubiquitous, convenient, on-
demand network access to a shared pool of
configurable computing resources (e.g., networks,
servers, storage, applications, and services) that can
be rapidly provisioned and released with minimal
management effort or service provider interaction.
Hardware devices, software, storage and network
infrastructure are made available to user through
Internet access. Rather than purchasing expensive
but powerful resources, users lease these resources
from the service providers. With cloud computing,
user can access the services via Internet access
regardless of time and location. They also get rid of
software installation in their local machine and able
to enjoy high availability of services. Furthermore,
high efficiency and fast deployment benefits are also
the attractions for company and individual who
moves to cloud services.
Due to the advantages such as flexibility,
mobility, and costs saving, the number of cloud user
has increased tremendously. Industry analysts have
made projections that entire computing industry will
be transformed into Cloud environment (Buyya et
al., 2009).
In this Cloud-driven era, security and privacy
concerns are becoming growing problems for the
user and the service provider. User authentication is
often the key issue in the Cloud environment. It is an
important operation for the service provider to verify
who can access their services and to identify the
group of each user.
Some commonly used authentication services
501
Wong K. and Kim M..
TOWARDS BIOMETRIC-BASED AUTHENTICATION FOR CLOUD COMPUTING.
DOI: 10.5220/0003909305010510
In Proceedings of the 2nd International Conference on Cloud Computing and Services Science (CLOSER-2012), pages 501-510
ISBN: 978-989-8565-05-1
Copyright
c
 2012 SCITEPRESS (Science and Technology Publications, Lda.)

include Kerberos (Neuman and Ts'o, 1994) and
OpenID (Recordon and Reed, 2006). The service
provider authenticates its users based on the
credential submitted such as password, token and
digital certificate. Unfortunately, these credentials
can often be stolen, accidentally revealed or hard to
remember. In view of this, we propose a biometric-
based authentication protocol that can be used as the
second factor for the cloud users to send their
authentication requests. Biometric authentication can
improve the quality of authentication (QοA) in cloud
environment. Our solution ensures both security in
the authentication and the privacy protection for all
sensitive information.
1.1 Organization
The rest of this paper is organized as follows: The
background for this research is in Section 2 and the
technical preliminaries are described in Section 3.
We present our proposed solution in Section 4
followed by the analysis in Section 5. Our
conclusion is in Section 6.
2 BACKGROUND
2.1 Cloud Computing Models
Cloud services are delivered in three fundamental
models (Lenk et al., 2009): Infrastructure as a
Service (IaaS), Platform as a Service (PaaS), and
Software as a Service (SaaS). IaaS is the lowest
level which is closest to the hardware devices
whereas, SaaS is the highest level that provides
services to the end-users. The Amazon web service
is one type of IaaS which has been widely used since
2006 while the Salesforce.com CRM system is an
example of SaaS.
PaaS level provides an application platform in
the cloud. Windows Azure platform is one example
of PaaS which enable the developers to build, host
and scale their applications in the Microsoft data
centers. Recently, a new concept called “Everything
as a Service (XaaS)” has been adopted as the new
trend in cloud computing. Several vendors such as
Microsoft and Hewlett Packard (Fiveash, 2008) have
been associated with it.
Biometric Authentication as a Service (BioAaaS)
has been defined as an approach for strong
authentication in web environments based on the
SaaS model (Senk and Dotzler, 2011).
2.2 User Authentication
When performing authentication over the Internet,
credential will be submitted by the principal (the
user, machine, or service requesting access)
(Convery, 2007). If the credentials match, the user is
allowed to access the services it subscribed from the
service providers. In this paper, we only consider
user as the principal who submits its credential for
authentication over the cloud.
There are several types of credential the users
can submit as proof of their identity. Shared-key is
typically password used protocols such as Password
Authentication Protocol (PAP) (Lloyd and Simpson,
1992) and Challenge Handshake Authentication
Protocol (CHAP) (Simpson, 1996).
Digital certificate is second type of credential
which can provide strong authentication in the cloud
environment. It is an electronic document which
uses a trusted Certificate Authority (CA) to blind the
encryption key with an identity (Canetti, 2004).
Decryption key is the only way to validate the
signed certificate.
Another type of credential is the commonly used
one-time-password (OTP) (Haller, 1994, Rubin,
1995). The end-user obtains the OTP from the token
(hardware or software) during the login time. The
token can generate a randomized password string
based on a complex algorithm in real time. Since the
password generated is unique and can only be used
once, OTP is possible to be used in the Cloud
environment. For example, Amazon Web Services
(AMW) has already started to use its OTP token for
use with individual AWS accounts (Brooks, 2009).
Recently, a German company BioID proposes
the world’s first biometric authentication service for
cloud computing (Krowneva, 2011). In their
solution, biometric authentication as a service
(BaaS) has been proposed to provide single sign-on
for user authentication.
2.3 Biometric-based Authentication
Biometric characteristics such as iris patterns, face,
fingerprints, palm prints and voice will be submitted
by the user as the credential for authentication over
the cloud. Biometric-based authentication systems
provide a higher degree of security as compared
with conventional authentication systems.
Furthermore, it allows the system to keep track of
the user’s activities because individual biometric
characteristics cannot be shared with others.
Generally, biometric authentication systems
consist of five modules, namely, the biometric
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
502

sensor, feature extractor, template storage, matching
module, and the decision module. Figure 1 illustrates
the general design for the biometric-based
authentication systems.
Figure 1: General design for biometric-based authenti-
cation systems.
During the enrolment process, the biometric
sensor scans the biometric traits of the user while the
feature extractor extracts the feature vector from the
scanned biometric data. The feature vector is then
stored in the template storage.
At the verification stage, the biometric sensor
and the feature extractor perform the same tasks as
in the enrolment process. However, the extracted
feature vector (query feature vector) will not be
stored in the storage. Instead, it will be used by the
matching module to compare with the templates
stored in the storage and output a similarity score.
The decision module is responsible in making
the final decision (accept or reject), which depends
on the similarity score or the threshold determined
by the system administrator.
3 TECHNICAL PRELIMINARIES
In this section, we describe some technical
preliminaries for our protocol design.
3.1 Security and Privacy Definitions
Security definition: In a generic sense, security is the
prevention of unauthorized party from gaining
access to confidential information and system
resources. A secure authentication system needs to
ensure that users are the persons they claim to be.
Our protocol is secure if no adversary party (the
malicious client, the malicious service provider or
the network intruder) gains access to the sensitive
information such as the verification code and the
template for each user, the shuffle protocol, all
decryption keys, and the original feature vector of
the user. During the authentication process, the
protocol must prevent the malicious service provider
from reconstructing the original feature vector based
on the verification code and the template stored in
the cloud. Also, the network intruder who watches
the traffic on the network must not learn anything.
Privacy definition: Information or data privacy is
referring to the ability of an individual or system to
prevent the leakage of any sensitive information to
any unauthorized party. A privacy-preserved system
should ensure that unauthorized party does not
improperly access confidential information.
In this paper, our solution bases on the user’s
biometric feature vector. Hence, we only consider
the privacy issues on the biometric template and the
verification code protections. No intermediate result
should leak any sensitive information and the service
provider cannot distinguish whether two
authentication requests belong to the same user.
3.2 Homomorphic Encryption Scheme
We will use the additive property of the
homomorphic encryption scheme proposed by
Paillier (Paillier, 1999) in our protocol.
Let
1
()
a
Em denote the encryption of message
1
m
with encryption key,
a
E . The scheme supports
the following operations in an encrypted form:
 Addition: Given two ciphertexts
1
()
a
Em and
2
()
a
Em , there exists an efficient algorithm
h
+
to compute
12
()
a
Em m+ .
 Scalar multiplication: Given a constant c and
a ciphertext
1
()
a
Em, there exists an efficient
algorithm
h
⋅
to compute
1
()
a
Ecm⋅ .
Note that when a scheme supports the additive
operation, it also supports scalar multiplication
because
1
()
a
Ecm
⋅
can be achieved by summing
1
()
a
Em successively
c
times.
By using the homomorphic encryption scheme,
we can compute the additive operation directly on
the encrypted data without the decryption. This is a
useful feature because the biometric template stored
in the server does not require decryption during the
matching operation. In our solution, the encryption
keys used by both parties are symmetric-key.
3.3 Notations Used
In Table 1, we summarize all the notations used
hereafter in this paper.
TOWARDSBIOMETRIC-BASEDAUTHENTICATIONFORCLOUDCOMPUTING
503

Table 1: Common notations used.
Notation Definition
X
original feature vector extracted from the
user during the enrolment process
Y
original feature vector extracted from the
user during the verification process
'X
transformed vector during the enrolment
process
'Y
transformed vector during the verification
process
''X
shuffled vector during the enrolment
process
''Y
shuffled vector during the verification
process
u
π
shuffle protocol for the user
U
'
i
x
i
-th element of
'X
'
i
y
i
-th element of
'Y
s
squared Euclidean distance
n
length of the original feature vector
m
length of the verification code
k
length of the transformed vector where,
4knm=+ +
TID
template identification number
VID
verification code identification number
u
E
encryption key from the user
U
u
D
decryption key from the user
U
p
E
encryption key from the service provider
p
D
decryption key from the service provider
()
pk
E ⋅
encryption operation by using the
p
k
E
()
pk
D ⋅
decryption operation by using the
p
k
D
ω
random non-zero number
Note that all datasets (both the original and the
transformed feature vectors) are ordered set.
4 PROPOSED SOLUTION
In our solution, the credential submitted by the user
consists of two parts: user’s biometric feature vector
and the verification code. Both parts must be
combined, transformed, and shuffled correctly in
order for the user to successful authenticate.
Like most existing biometric-based
authentication systems, our solution consists of both
the enrolment and the verification processes. In the
following sections, we will describe in details the
components and the authentication workflows of our
solution.
Figure 2: Basic components for the proposed solution.
4.1 Components
We now formally describe the components in our
proposed solution as follow: (as illustrated in Figure.
2):
 User: individual who sends the authentication
request.
 Client: computer or workstation with Internet
access.
 Service provider: company or organization
who provides cloud services (SaaS, PaaS or
IaaS) to the user.
 Template (T): transformed feature vector
stored in the cloud storage.
 Query feature vector (Q): transformed feature
vector used to compare with the template.
Hereafter in this section, we refer sensitive
information as our protected information includes
the biometric template, the verification code, and the
shuffle protocol.
Both the client and the service provider manage
separate components in order to verify the user. The
client has the following components:
 Biometric sensor: scans the biometric traits of
the user.
 Feature extractor: extracts the feature vector
from the scanned biometric data.
 Verification code generator: generates unique
verification code for the user.
 Transformation module: transforms the
original feature vector and shuffles the
transformed feature vector.
 Encryption module: encrypts the transformed
and shuffled feature vector with the correct
encryption key (i.e., encrypts with the user’s
key during the enrolment process).
 Decryption module: decrypts the computation
output.
The service provider requires the following
components:
 Verification code retrieval: retrieves the
verification code for the user.
 Templates storage: stores the template for
each user.
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
504

X
(')
u
E
V
V
''
X
( '')
u
E
X
{,('')}
u
TID E X
{,(')}
u
VID E V
Figure 3: The overview of the enrolment process.
 Verification codes storage: stores the
verification code for each user.
 Computation module: performs the squared
Euclidean distance (
s
) computation between
the query feature vector and the template.
 Decision module: making the final decision by
comparing the
s
with the given threshold
τ
.
4.2 Enrolment
The objective of the enrolment process is to process
the scanned biometric data and extract a set of
feature vector to be stored as the template for the
user. The enrolment process is required for the new
user who wants to join the cloud. A successful
enrolment process enables the user to receive the
TID
and the
VID
.
4.2.1 Transformation
Let
{}
12
, ,...,
n
Xxx x=
,
0n >
and
{
}
12
, ,...,
m
Vvv v=
,
0m >
be the feature vector of
the user and the verification code generated,
respectively. We transform
X
into
{}
'
' | 1, 2,..., 4
i
Xxi nm== ++
such that
'
ii
x
x
=
for
1 in≤≤
,
'
nj j
x
v
+
=
for
1 jm≤≤ ,
''
12
1
nm nm
xx
++ ++
==
,
'2
31
n
nm i i
x
x
++ =
=
∑
and
'2
41
m
nm j j
x
v
++ =
=
∑
.
4.2.2 Shuffle Protocol
We require a shuffle protocol (
u
π
) to permute the
order of elements in the transformed vector
'
X
. We
use the same shuffle protocol during the verification
process for the same user.
4.2.3 Overview of the Enrolment Process
We illustrate the overview of the enrolment process
in Figure. 3 and the workflow as follow:
1.
The biometric sensor scans the biometric trait
of the user.
2.
The feature extractor processes the scanned
biometric data to extract the feature vector of
the user,
12
{ , ,..., }
n
X
xx x
=
.
3.
The feature extractor sends
X
to the
transformation module.
4.
The verification code generator generates a
unique verification code
1
2
{, ,..., }
m
Vvv v=
for
the user and sends it to the transformation
module.
5.
We transform the verification code into
'2VV
=
− and encrypt it using the encryption
key from the user. Then, it will be stored at the
verification codes storage.
6.
The transformation module transforms
X
into
'
X
. Next, it shuffles the transformed vector
'
X
i.e. '' ( ')
u
X
X
π
=
before sending it to the
encryption module.
7.
The encryption module encrypts
''
X
with the
user’s encryption key. Finally, the
('')
u
EX is
sent to the service provider and stored as the
user’s template in the templates storage.
4.3 Verification
When the user wants to access data stored in the
cloud storages or uses the cloud services, the user
must be authenticated first. The verification process
is responsible to verify the users who they claim to
be.
TOWARDSBIOMETRIC-BASEDAUTHENTICATIONFORCLOUDCOMPUTING
505

Y
''Y
( '')
p
E
Y
TID
()
u
E
s ω
⋅
( '')
p
E
Y
()
p
E
s ω
⋅
(')
u
E
V
VID
Figure 4: The overview of the verification process.
4.3.1 Transformation
Let
{
}
12
, ,...,
n
Yyy y=
,
0n >
and
{
}
12
, ,...,
m
Vvv v=
,
0m >
be the feature vector
extracted from the user and the verification code,
respectively. The verification code used must be the
same in both enrolment and verification processes.
We transform
Y
into
{}
'
' | 1, 2,..., 4
i
Yyi nm
=
=++
such that
'
2
ii
yy=−
for
1 in≤≤
,
'
2
nj j
yv
+
=−
for
1 jm≤≤
,
'2
11
n
nm i i
yy
++ =
=
∑
,
'2
21
m
nm j j
yv
++ =
=
∑
,
''
34
1
nm nm
yy
++ ++
==
. The length for
'Y
must be same
as
'
X
which is
4knm=+ +
.
4.3.2 Shuffle Protocol
We require the same shuffle protocol used in the
enrolment process during the verification process.
The transformed feature vector
'Y
needs to be
shuffled in the same order as
'
X
.
4.3.3 Overview of the Verification Process
The workflow for the verification process is as
follow (as illustrated in Figure. 4):
1.
The biometric sensor scans the biometric trait
of the user.
2.
The feature extractor processes the scanned
biometric data to extract the feature vector of
the user,
12
{ , ,..., }
n
Yyy y= .
3.
The feature extractor sends
Y
to the
transformation module.
4.
Next, the client retrieves the verification code
by using the
VID
.
5.
The verification code retrieval retrieves the
verification code for the user from the storage
which is associated with the
VID
.
6.
The transformation module receives the
verification code
(')
u
EV
.
7.
The transformation module computes
(
)
(')
uu
D
EV
and transforms
Y
into vector
'Y
.
Next, it shuffles
'Y
i.e. '' ( ')
u
YY
π
= and sends
''Y
to the encryption module.
8.
The encryption module encrypts
''Y
with the
service provider’s encryption key
p
E
. Next, the
('')
p
E
Y
is sent together with the TID to the
computation module.
9.
The computation module retrieves ('')
u
EX
from the templates storage which is associated
with the
TID .
10.
If both ('')
u
EX and
('')
p
E
Y
have the same
size, the computation module computes:
i.
Decryption:
(
)
( '') ''
pp
DEY Y=
ii.
Scalar multiplication:
'' ( '') ( '' '')
uu
YEX EXY
⋅
=⋅
iii.
Homomorphic additive operation:
()
(
)
4'' ''
1
()
nm
uuiii
Es E x y
++
=
=⋅
∑
iv.
Add noise: () ( )
uu
Es E s
ω
ω ⋅=⋅,
where
ω
is a random non-zero
number.
The computation module sends ()
u
Es
ω
⋅ to the
client.
11.
The decryption module of the client decrypts
()
u
Es
ω
⋅
and then encrypts
s
ω
⋅ with
p
E
.
Then, the decryption module sends
()
p
E
s
ω
⋅
to
the decision module of the service provider for
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
506

making the decision. The decision module
decrypts
()
p
E
s
ω
⋅
and makes the decision as
follows (
τ
is the threshold determined by the
service provider):
,
,
Accept if s
decision
Reject if s
τ
τ
<
⎧
=
⎨
>
⎩
Note that for different authentication requests,
we may require different security levels. Hence,
our system can assign different threshold values
for different users.
5 ANALYSIS
In this section, we present the correctness, security,
privacy and efficiency analysis for our proposed
solution.
5.1 Correctness Analysis
Theorem 1. Our protocol correctly computes the
squared Euclidean distance between the query
feature vector and the template stored in the storage
if both the client and the service provider follow the
protocol faithfully.
Proof. Let
12
{ , ,..., }
n
X
xx x= be the extracted
feature vector of user
A
during the enrolment
process. We transform the feature vector
X
into
'
X
:
()()
11
22
11
,..., , ,..., ,1,1,
'
,
nm
nm
ii jj
xxvv
X
x
v
==
⎧⎫
⎪⎪
=
⎨⎬
⎪⎪
⎩⎭
∑∑
(1)
Then, we randomly shuffle the order of elements
in
'
X
. Let '' ( ')
A
X
X
π
= be the shuffled vector by
using the shuffle protocol
A
π
. Next, we encrypt
''
X
by using the encryption key
A
E and store the
following result as the template in the templates
storage:
()()
1
1
22
11
( ),..., ( ),
( ),..., ( ),
('')
(1), (1),
,
AAn
AAm
A
AA
nm
AiiAij
Ex Ex
Ev Ev
EX
EE
ExEv
==
⎧⎫
⎪⎪
⎪⎪
=
⎨⎬
⎪⎪
⎪⎪
⎩⎭
∑∑
(2)
Note that for ease of explanation, we do not change
the order of elements in Eq. (2).
Assume that
12
{, ,..., }
n
Yyy y= is the query
feature vector during the verification process. The
client retrieves the verification code from the service
provider and transforms
Y
into
'Y
as follows:
()()
11
22
11
2 ,..., 2 , 2 ,..., 2 ,
'
,,1,1
nm
nm
ii jj
yyvv
Y
yv
==
−−−−
⎧
⎫
⎪
⎪
=
⎨
⎬
⎪
⎪
⎩⎭
∑∑
(3)
By using the same shuffle protocol
A
π
(if the
user is
A
), the client computes '' ( ')
A
YY
π
= and
encrypts
''Y
with the encryption key
P
E to produce:
()()
1
1
22
11
( 2 ),..., ( 2 ),
( 2 ),..., ( 2 ),
('')
,,
(1), (1)
PPn
PPm
P
nm
PiiPij
PP
EyEy
Ev Ev
EY
ExEv
EE
==
−−
⎧
⎫
⎪
⎪
−−
⎪
⎪
=
⎨
⎬
⎪
⎪
⎪
⎪
⎩⎭
∑∑
(4)
For ease of explanation, we do not change the order
of elements in Eq. (4).
The squared Euclidean distance is computed as
follow: The service provider first decrypts
( '')
P
EY
to obtain
''Y
and computes the scalar multiplication
for each
i
-th element in
''Y
and ( '')
A
EX according
to their index position:
'' ( '')
A
YEX
⋅
('''')
A
EXY
=
⋅
(
)
(
)
()( )
()()
()
()
()
()
11
11
22
11
22
11
2 ( ) ,..., 2 ( ) ,
2 ( ) ,..., 2 ( ) ,
(1), (1),
1,1
AnAn
AmAm
nm
ii A jj A
nm
Aii Aij
yEx y Ex
vEv v Ev
yE vE
ExEv
==
==
⎧
⎫
−⋅ −⋅
⎪
⎪
−⋅ − ⋅
⎪
⎪
⎪
⎪
=
⎨
⎬
⋅⋅
⎪
⎪
⎪
⎪
⋅⋅
⎪
⎪
⎩⎭
∑∑
∑∑
()()
()()
11
22
1
22
11
22
11
(2 ),..., (2 ),
(2 ),..., (2 ),
,,
,
AAnn
AAm
nm
AiiA jj
nm
AiiAij
Exy Exy
Ev Ev
EyEv
ExEv
==
==
−−
⎧
⎫
⎪
⎪
−−
⎪
⎪
=
⎨
⎬
⎪
⎪
⎪
⎪
⎩⎭
∑∑
∑∑
(5)
Next, the service provider computes
homomorphic additive operation for each
(
)
()
'' ''
'' ''
ii
x
yXY⋅∈ ⋅
in Eq. (5):
(
)( )
2
11
() 2 2
nm
AAiiihAjj
Es E xy E v
==
=−+ −
∑∑
(
)
(
)
()()
22
11
22
11
nm
hA i i h A j j
nm
hA i i h A i j
EyEv
ExEv
==
==
++
++
∑
∑
∑∑
(
)( )
2
11
2
nn
AiihAi ii
ExE xy
==
=+−
∑∑
(
)
2
1
n
hA i i
Ey
=
+
∑
(
)
(
)
22
1
2
n
Aii iii
Exxyy
=
=−+
∑
()
(
)
2
1
n
Aiii
Exy
=
=−
∑
(6)
TOWARDSBIOMETRIC-BASEDAUTHENTICATIONFORCLOUDCOMPUTING
507

After we decipher the result in Eq. (6), we can
obtain the squared Euclidean distance
2
1
()
n
iii
s
xy
=
=−
∑
. Note that in Eq. (6), we
eliminate the verification code and all additional
features. Hence, if the service provider retrieves the
correct verification code and the client computes
''Y
correctly, our protocol outputs the correct
squared Euclidean distance for
X
and
Y
.
Theorem 2. If one of the parties (either the client or
the service provider) is not following the protocol,
the final output will not reflect the squared
Euclidean distance for the two vectors (
X
and
Y
).
Subsequently, the verification process will fail and
the user cannot access the system.
Proof. The client or the service provider who is not
following the protocol is considering as the
malicious party in our protocol. The proof of this
theorem is same as the proof in Theorem 3 and
Theorem 4 under the security analysis.
5.2 Security Analysis
In this section, we will analyze two possible attacks:
internal and external attack. Internal attack involves
malicious party such as employee at client who
attempts to gain access into the cloud. External
attack involves external parties (intruders or network
attackers) who watch the traffic on the network.
They are interested in learning some knowledge
from the computation protocol or intercept the data
in the network. Note that internal attack is more
serious as compared to the external attack because
attackers are having more knowledge about the
protocol.
Theorem 3. Our protocol is secure against malicious
user who tries to gain access to the cloud. Without
the knowledge of sensitive information and the
decryption key, the authentication is not possible for
attacker at the client side.
Proof. During the enrolment process, the system
generates the biometric template for each user. Only
the user who enrolled into the cloud has its template
and the verification code stored in the cloud
storages. In the absence of the template, the system
cannot authenticate the user.
In our protocol, any malicious user who wants to
pose as an enrolled user must gain access to three
sensitive information: (1) the verification code, (2)
the original feature vector and (3) the shuffle
protocol. Since the verification code is stored at the
cloud storage and is encrypted using the encryption
key of the respective user, the attacker is not able to
view it because he has no knowledge about the
decryption key. If the attacker gains access to the
original feature vector of the user, he is not able to
use it directly for the verification process because
the verification code and the shuffle protocol are not
accessible. In the worst scenario, if the attacker
obtains the decryption key of any user, the security
for the user is still can be guaranteed. Hence, our
protocol is secure against attacker who tries to gain
access to the cloud system.
Theorem 4. Our protocol is secure against malicious
service provider who tries to gain access to the
verification codes and templates stored in the cloud
storages. The malicious service provider is not able
to reconstruct the original feature vector of any user.
Proof. A malicious service provider wants to learn
the original feature vector of the user. Although the
verification code and the template are reside in the
service provider’s side, but the malicious service
provider is not able to reconstruct the original
feature vector for any user. This is because the
verification code and the template are encrypted
using the encryption key from each respective user.
The service provider has no knowledge about the
decryption key of any user. Gaining access to these
encrypted vector is as difficult as attacking the
encryption algorithm. Brute-force attack is also
impossible since all the verification codes and the
templates are different (after the encryption
operation). Hence, our protocol is able to prevent the
malicious service provider from reconstruct the
original feature vector of the user.
Theorem 5. Network attacker who listens to the
traffic is not able to learn any sensitive information.
Proof. In our protocol, all the data transmit over the
network (between the client and the service
provider) are encrypt either with the user’s
encryption key or with the service provider’s key.
When the network attacker watches the network, he
cannot learn any information because he has no
knowledge about the decryption key. During the
verification process, network attacker is not possible
to be authenticated by the cloud because he has no
knowledge about any sensitive information. Hence,
our protocol is secure against the network attacker.
5.3 Privacy Analysis
The main privacy concern in our protocol is the
amount of information revealed to the service
provider during the authentication process. Our
protocol should ensure the confidentiality of all
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
508

sensitive information. All the intermediate results
and the final output will not compromise the privacy
of the user.
Theorem 6. The feature vector extracted from the
user is never stored in their original form. Although
the template and the verification code are stored
with the service provider, but the service provider
learns nothing from its storage.
Proof. In our protocol, we encrypt both the
verification code and the template using the
encryption key from the user and then store the
encrypted data at the service provider's side. The
service provider is not able to learn anything
because it has no knowledge about the decryption
key from the user. In the worst scenario, if the
decryption key of the user has been compromised,
the service provider also not able to identify the
original feature vector of the user because the
template has been transformed and shuffled during
the enrolment process. Furthermore, we include
some additional features and the verification code
into the template.
During the verification process, the service
provider decrypts the
( '')
P
EY before performing the
scalar multiplication operation. After the decryption,
the service provider is not able to identify the
original feature vector and verification code used in
the query feature vector. Hence, our protocol
protects both the verification code and the template
stored in the cloud storages.
Theorem 7. The service provider is not able to
distinguish whether two authentication requests
belong to the same user.
Proof. In our protocol, the verification code and the
template are stored separately in the service
provider’s side. This design prevents the malicious
party from knowing which verification code is
associated with which template in the case when
both storages are compromised. The decision
module makes the verification decision based on the
similarity score (squared Euclidean distance) and the
threshold determined by the system. If the similarity
score is lower than the threshold, it can reject the
user. Otherwise, the system verifies the user and the
authentication process is successful. With only the
similarity score, the decision module is not able to
distinguish whether two authentication requests
belong to the same user.
5.4 Efficiency Analysis
The total communication cost depends on the
amount of data transferred in the protocol. During
the verification process, the client sends one query
feature vector with 4knm
=
++encrypted data to
the service provider. The service provider responds
with one encrypted result. The client then replies
with one encrypted value. The communication
complexity incurred by our protocol is
()Ok
.
In terms of complexity, our protocol requires
()Ok
encryptions,
()Ok
scalar multiplications and
()Ok
homomorphic additive operations.
6 CONCLUSIONS
In this paper, we proposed a biometric-based
authentication protocol for cloud computing. Our
target is to achieve secure authentication while
protecting the sensitive information of users. We
incorporate the homomorphic encryption scheme in
our Squared Euclidean distance computation that
allows us to compare both the query feature vector
and the template in an encrypted form. In order for
the user to successful authenticate the biometric
feature vector and the verification code must be
correctly transformed and shuffled. Our solution
preserves the privacy for sensitive information of
users and securely performs the authentication
process.
REFERENCES
Brooks, C. 2009. Amazon adds onetime password token to
entice the wary. SearchCloudComputing.
Buyya, R., Yeo, C. S., Venugopal, S., Broberg, J. &
Brandic, I. 2009. Cloud computing and emerging IT
platforms: Vision, hype, and reality for delivering
computing as the 5th utility. Future Gener. Comput.
Syst., 25, 599-616.
Canetti, R. 2004. Universally Composable Signature,
Certification, and Authentication. Proceedings of the
17th IEEE workshop on Computer Security
Foundations. IEEE Computer Society.
Convery, S. 2007. Network Authentication, Authorization,
and Accounting Part One: Concepts, Elements, and
Approaches. The Internet Protocol Journal, 10, 2-11.
Fiveash, K. 2008. HP sells cloud vision amidst economic
downpour. Will customers get soaked on
transformation journeys? : King's College London.
Haller, N. 1994. The S/KEY One-Time Password System.
Internet Society Symposium on Network and
Distributed Systems.
Krowneva. 2011. BioID Announces World's First
Biometric Authentication as a Service (BaaS)
[Online]. Available: http://silicontrust.wordpress.com/
TOWARDSBIOMETRIC-BASEDAUTHENTICATIONFORCLOUDCOMPUTING
509

2011/03/04/bioid-announces-worlds-first-biometric-
authentication-as-a-service-baas/ [Accessed].
Lenk, A., Klems, M., Nimis, J., Tai, S. & Sandholm, T.
2009. What's inside the Cloud? An architectural map
of the Cloud landscape. Proceedings of the 2009 ICSE
Workshop on Software Engineering Challenges of
Cloud Computing. IEEE Computer Society.
Lloyd, B. & Simpson, W. 1992. PPP Authentication
Protocols. RFC Editor.
Mell, P. & Grance, T. 2009. The NIST Definition of
Cloud Computing. Available: http://
www.csrc.nist.gov/groups/SNS/cloud-
computing/cloud-def-v15.doc.
Neuman, B. C. & Ts'o, T. 1994. Kerberos: An
Authentication Service for Open Network Systems.
IEEE Communications, 32, 33-38.
Paillier, P. 1999. Public-key cryptosystems based on
composite degree residuosity classes. Proceedings of
the 17th international conference on Theory and
application of cryptographic techniques. Prague,
Czech Republic: Springer-Verlag.
Recordon, D. & Reed, D. 2006. OpenID 2.0: a platform
for user-centric identity management. Proceedings of
the second ACM workshop on Digital identity
management. Alexandria, Virginia, USA: ACM.
Rubin, A. D. 1995. Independent one-time passwords.
Proceedings of the 5th conference on USENIX UNIX
Security Symposium - Volume 5. Salt Lake City, Utah:
USENIX Association.
Senk, C. & Dotzler, F. 2011. Biometric Authentication as
a Service for Enterprise Identity Management
Deployment: A Data Protection Perspective. Sixth
International Conference on Availability, Reliability
and Security. Vienna Austria.
Simpson, W. 1996. PPP Challenge Handshake
Authentication Protocol (CHAP). RFC Editor.
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
510