Ingénierie des Systèmes d’Information

Vol. 24, No. 3, June, 2019, pp. 343-351

Journal homepage: http://iieta.org/journals/isi

A Hybrid Privacy Preserving Scheme Using Finger Print Detection in Cloud Environment
Garima Verma1*, Rupak Chakraborty2
1
DIT University, Dehradun, India
2
Bennett University, Greater Noida, India

Corresponding Author Email: [email protected]

https://doi.org/10.18280/isi.240315 ABSTRACT

Received: 6 April 2019 Cloud computing provides a new paradigm of computing. It offers a scalable, manageable and
Accepted: 10 June 2019 huge pool of resources that can be accessed by users from anywhere anytime. It also ensures
the integrity of data stored on the cloud. But ensuring the confidentiality and integrity of
Keywords: sensitive information is still a big challenge. To overcome this challenge, a hybrid two-phase
cloud computing, security, biometric, security system for preserving the privacy of data on the cloud has been proposed. The hybrid
fingerprint detection, minutiae points, approach combines feature extraction and encryption techniques to enhance the security of
elliptic curve accessing data from the cloud. At first, the minutiae point has been extracted from the
biometric fingerprint, locally collected from the state university in Northern India. The private
key has been finalized by generating an elliptic curve using the minutiae point for achieving
better encryption of fingerprint. The effectiveness of the approach has been tested in terms of
similarity score, False Matching Ratio (FMR), False Non Matching Ratio (FNMR) and
recognition accuracy, when applied on the local fingerprint database. The evidence of the
outcomes suggests that the proposed technique ensures relatively improved security and
privacy of data in the cloud system as compared to some recent state-of-art methods.

1. INTRODUCTION (2) Techniques are diminutive chances of stealing.

(3) Techniques are cost effective and accurate.
Biometrics identification is one of the most popular Techniques are easy, user-friendly, and secure [4], [5].
methods used nowadays for identifying the authenticity of an Nowadays, many of the smart devices like phones, laptops,
individual. All methods of biometric identification such as doors, etc. are using an authentication mechanism based on
face recognition, iris, etc. have their own uniqueness [1], [2], biometric techniques instead of a simple password or swap
such as two persons cannot have same fingerprint, and cards or token system [4]. There are very fewer chances to
persistence. As biometric features usually do not change over break the system unlike other traditional methods because
time and age, so biometric-based recognition systems are every individual has unique biometric features and patterns [5]
being focused day by day for identification and security of data Due to all the above reasons, biometric authentication systems
on the cloud. These biometric systems perform user are reliable and suitable for cloud access also. The biometric
authentication by verifying an individual's characteristics. For data of all the users who access the cloud can be stored and
this, it is required to maintain the database of biometric verified at the time of cloud utilization. The safety of data,
features of all individuals. Whenever any user wants to access especially sensitive data like biometric data or other data and
data or resources, first, the process of verification starts. In the managing privacy preservation are the biggest challenge in the
verification process, a user’s biometric features are matched cloud computing systems [2]. The popularity of cloud
with the stored template in the database using any matching computing has forced researchers and developers to handle
technique [3]. this issue very carefully. This can be achieved by the
All biometric methods are generally categorized into two encryption of data available on the cloud, importantly
categories 1) Physical, 2) Behavioral [4]. Physical biometric biometric data, which will provide better security and privacy
methods are a fingerprint, palm print, iris identification, retinal protection [6]. In this paper, a novel biometric-based system
scanning, face recognition, etc. while the behavioral biometric using a fingerprint detection technique has been proposed for
methods are DNA matching, voice recognition, signature, better privacy preservation and security in cloud systems.
handwriting, etc. [4]. All type of biometric characteristics is Biometric image templates are encrypted using the elliptic
unique and measurable for identification and verification of an curve with the digital signature encryption algorithm. For
individual [5]. There are various advantages of using providing better security and privacy on the cloud system the
biometric authentication as compared to conventional paillier algorithm has been used. The main advantage of using
techniques of verification like cryptography. Some of the the paillier algorithm is its Homomorphic encryption
advantages of using biometric authentication are as listed properties [7]. The road map of this paper is presented as
below: follows. Section 2 discusses the related work done in the past
(1) Biometric methods give the category of authentication by various researchers. Section 3 presents the proposed work.
called as something you have. A person need not remember or Section 4 describes the results generated from experiments and
carry identification separately like smart card, password, etc. at last section 5 draws the conclusion of the work.

343
2. LITERATURE REVIEW biometric systems. Zhang et al. [16] proposed a new privacy-
preserving scheme based on biometric identification which
A lot of research has been done in the past to make secure ensures lightweight database computations. They have
cloud computing systems using various techniques. In most of designed a biometric data encryption algorithm and introduces
the time, researchers have used the traditional cryptography perturb terms in biometric data. The biggest challenge in cloud
techniques for providing security and privacy of data in the systems is to provide an efficient solution for security that
cloud. The main hassles with these techniques were in gives access to resources and data which are outsourced to the
handling of security keys and data. For example, if the cloud. To overcome this issue, Kumari et al. [17], devised a
passwords are used for authentication of users then he may biometric authentication system for the multi-cloud server.
have a problem of remembrance. Especially if a user has They have used the bio-hashing technique for better accuracy
several types of accounts then setting many passwords and of pattern matching. Al et al. [18], addressed security issues of
remembering all these passwords is a hard task. Some other mobile cloud computing by presenting an effective model to
situations may arise like, if a user puts the same password for solve the identification problem in the mobile cloud using
all his accounts, then it will provide a possibility of hacking all fingerprints. They have combined fingerprints with a
accounts. If the password is hacked or if the user saves the password to make the system much strong. Shakil et al. [19],
password in some file, then all accounts will be hacked if that proposed the biometric authentication system for the health
file is hacked. To avoiding the situation of remembrance of care database by introducing a signature-based system. with
password, smart cards can be used but, which have to be the help of a back-propagation network. Encouraged by the
carried by the user all the time. If anytime it is Lost or stolen, stated techniques, one hybrid approach in combination with
then it may push users to some critical situations that can be the biometric and encryption technique has been proposed to
considered as a major drawback of using smart cards. The preserve better security as well as privacy in the cloud system.
above stated problems can be solved up to a great extent with The number of pages for the manuscript must be no more
biometric authentication due to its most important property i.e. than ten, including all the sections. Please make sure that the
“something that you have”. whole text ends on an even page. Please do not insert page
Literature reveals that Bhattasali et al, [8] surveyed various numbers. Please do not use the Headers or the Footers because
biometric techniques in their work. Authors claimed that they are reserved for the technical editing by editors.
remote accessing of any type of data using biometric systems
is more challenging in comparison to access from a local place. 2.1 Author’s contribution
In these situations, it is unavoidable to prevent unauthorized
access. Biometric authentication systems are more efficient in The detailed contribution of the work is as follows:
comparison to the traditional system of authentications. (1) The work uses biometric authentication via fingerprint
Naveed et al, [9] analyzed the various biometric authentication detection with PCA, elliptic curve encryption and
techniques in the cloud computing environment and explores homomorphic encryption using the Paillier algorithm.
how these techniques could help in reducing security threats. (2) It has introduced a strong user authentication method as
The privacy reserving cloud-based system with biometric well as overcome the problems that can occur due to
identification has been proposed by Haghighat et al, [10]. traditional cryptographic authentication methods for user
Authors have used k-d tree approach to create encrypted authentication.
queries for preserving data secure. In the year 2016, Hahn et (3) The main advantage of the proposed system is more
al, [11] proposed an effective privacy preserving fingerprint secure, fast, less power consumption and fewer chances of data
identification scheme for cloud computing systems with a leakage due to the use of elliptic curve encryption.
homomorphic encryption scheme. The authors tested the (4) For proving the effectiveness of the work, quantitative
proposed scheme on the Amazon EC2 cloud. In the year 2018, comparisons have been carried out, in terms of FMR and
Bala et al, [12] presented a biometric-based homomorphic FNMR, accuracy and recognition rate.
encryption algorithm for data transmission in cloud systems. The next section describes the proposed system and its
The proposed scheme was able to handle phishing and working along with the block diagram.
shoulder surfing attacks in the cloud environment. In a study
done by Pan et al, [13] authors said that biometric
identification provides lots of convenience to users of cloud 3. PROPOSED SYSTEM
computing systems but simultaneously increases privacy
concerns also. In this study, researchers have studied various Two steps will be used in the proposed system for providing
attacks and also validated them in a cloud environment. Kumar secure access- 1. Enrollment of fingerprint, and 2. Verification
et al. [14], proposed a security scheme using face recognition of Fingerprints. In the proposed system, fingerprint biometric-
biometric identification approach in their proposed scheme on based identification of individual users will be used. The main
the cloud computing environment. As the main focus of the reasons for considering fingerprint as biometric for
proposed work is on cloud security and privacy, so literature identification are the advantages it offers in comparison to
survey of security-oriented research papers has been continued. other biometrics. For example, no two fingerprints are the
Lee et al. [15], analyzed the benefits of fingerprint same, it does not change with age, small storage is required in
identification in comparison to other biometric forms. The comparison to other biometrics, devices are comparatively
author has also discussed various case studies of companies in cheap, easy to use, and require low maintenance cost [14-17].
the UK, to justify his work and proved that the fingerprint The block diagram of the proposed system is shown in Figure.
identification system is comparatively better than other 1.

344
 Enrollment Phase
Preprocessing
Input Fingerprint
Fingerprint Machine

Feature Template
extraction Generation

Cloud Database
Encryption

Verification Phase
Preprocessing
Query Fingerprint
Fingerprint Machine Yes
Similarity
Feature Template Recognized
Score If score=1
extraction Generation
Calculation
No

Not
Encryption Recognized

Figure 1. Proposed system block diagram

3.1 Enrollment phase

In the enrollment phase, shown in Figure. 2, an individual

user fingerprint is enrolled and stored in the database by
fingerprint detection machine. After the storage of an
individual's image, its quality is checked and if the quality
level is appropriate then feature extraction is done. The
proposed system uses the minutiae point algorithm [20] for
feature extraction. Minutiae points are very important and Figure 3. Ridge ending and ridge bifurcation
widely used features of the fingerprint detection technique.
These are used for matching an appropriate fingerprint with After the conversion of a binary image, it goes through
stored templates of fingerprints in the database. Minutiae morphological thinning, which reduces ridge to one pixel of
points are used to distinguish one fingerprint image from thickness for minutiae extraction. In the thinned binary image
others. A fingerprint image with good quality can have 25 to each pixel (p) is analyzed to find the location of minutiae. This
85 minutiae points [21]. These minutiae points are is achieved by Eq. 2, called a Rutovitz crossing number [22].
individualities in the finger ridge patterns of an individual. In
this, the two most widely used features are ridge ending and
ridge bifurcation. Ridge ending is the sudden end point of the (2)
ridge while ridge bifurcation is the point where two or more
branches are generated from the single ridge shown in Figure where val is 0 or 1. Minutiae locations ridge ending is now
3. For extraction of minutiae points the binary image-based identified as cn=1 and ridge bifurcations are cn=3.
method is used. This method requires to convert each There are four parameters that are used to describe each
grayscale pixel to the binary values 0 or 1. detected minutiae point. Let ith minutiae point is represented
as Mi.
1, 𝑖𝑓 (𝑥, 𝑦) ≥ 𝑡
𝐼(𝑥, 𝑦) = { (1)
0, 𝑂𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 Mi = (xi,yi,θi,Ti)

where minutiae point coordinates are xi and yi, θi is the

name of
template direction of inutiae, Ti is the called a type of minutiae point
whether ridge ending or bifurcation.
Check the Further, minutiae points are a most accepted feature in
quality of Feature
finger extraction
Database fingerprint but, the pre-processing of the image does not
image eliminate all the defects from the original grey level image.
For that, we require some other approach to remove false
minutiae points caused by poor quality. We have applied the
Principal Component Analysis (PCA) technique for improving
Figure 2. Enrollment phase the recognition rate of fingerprints [23].

345
3.1.1 Principal component analysis 3.1.3 Point multiplication
PCA is a statistical approach that translates an image of In this operation, a new point Q can be obtained on the same
MxN to a vector according to the rows and columns. Therefore, elliptic curve by multiplying a point P on the elliptic curve and
the image of size M x N becomes a vector of dimension (D). is with scalar k, kP=Q. To obtain point multiplication two
PCA is used as a tool for the reduction of multidimensional basic operations are used-
data to low dimensions. The basic steps of PCA are shown in – Point addition it is an operation of adding two points to
Fig. 6. The main advantage of using the PCA technique is the obtain another point.
reduction of complexity in images and also reduction of noise
level because of choosing a maximum variation basis, so the R = P + Q, if P 6= Q
background small variations are ignored automatically.
– Point doubling It is an operation in which a new point is
Begin obtained by adding a point P by itself.
1. Resized square images of 2N represents the set of M
images {I1, I2 ---- Im} R = 2P if P + Q then P + Q = 2P
1
2. Average of training set 𝜇 = ∑𝑀 𝑛=1 𝐼𝑛
𝑚
3. Vector W is different for each image in training set 𝑊𝑖 = 3.1.4 Point subtraction
𝐼𝑖 − 𝑢 Let P and Q are two distinct points, such that
4. Calculate Covariance matrix
5. Measure the eigenvectors and eigenvalues of the P = (xj, yj), Q = (xk, yk), then P−Q = P+(−Q), where Q =
covariance matrix
(xk,−yk mod p)
End
3.1.5 Multiplicative inverse
Figure 4. Steps of PCA
Let Fp is a finite field, if 𝑥 ∈ 𝐹𝑝 such that ax ≡ 1( mod p),
3.1.2 Encryption of fingerprint ax = 1 in Fp , x is called multiplicative inverse of a, can be
This is the second step in the proposed system for adding represented by a−1 mod p.
more security for individual user identification. In this Elliptic The main advantage of using elliptic curve cryptography is
curve encryption with a digital signature algorithm is applied that it uses smaller keys in comparison to other cryptographic
for encryption of fingerprint templates. It is public key algorithms for very fast key generation. At the time of
cryptography, which is based on the algebraic structure of enrollment of fingerprint, the system takes an input image
elliptic curve over finite fields. An elliptic curve is shown in from which features are extracted. After the extraction of the
Fig. 7 and equations of the elliptic curve are given as 3 and 4. features, the fingerprint template is encrypted using elliptical
encryption. The identification of the user is done by similarity
y2 = x3 + ax + b (3) matching between query and stored image templates. The
encrypted test fingerprint templates are matched with stored
database templates [24-25].
4a3 + 27b2 6= 0 (4) In the proposed method, the coefficients of the elliptic curve
are generated from minutiae points. Before the encryption
process, there is a need to generate private and public keys.
A pair (x,y), where , is a point on the elliptic curve,
Following are the steps to generate private and public keys
if (x,y) satisfies the Eq. 3. The point at infinity is also said to
shown in Algorithm 2 shown in Figure. 6. After the generation
be on the same curve. The set of all the points on the elliptic
of keys, a digital signature is generated. Using a digital
curve (E) is denoted by E (Fp). Let E be an elliptic curve
signature recipient of a message can verify the message
defined over a finite field Fp. Let P be a point in E (Fp), and
authenticity using a public key. For this, a secure hash
suppose that P has prime order n. Then the cyclic subgroup of
algorithm is used to convert the variable-length message into
E (Fp) generated by P is-
a fixed-length message called as digest h(m), signature
generation is shown as Algorithm -3, shown in Figure. 7. After
< P >= {∞, P, 2P, 3P ,......,(n − 1)P} (5)
the generation of the signature, it needs to be verified for the
query message for its authenticity using the public key.
The prime p, the equation of the elliptic curve E, and the
Signature verification algorithm in Algorithm -4, shown in
point P and its order n are the public domain parameters. There
Figure. 8. The system uses Paillier Homomorphic encryption
are three operations done on an elliptic curve.
for storing and accessing data in the cloud for more security
and privacy. If the user is authorized then he can access data
from the cloud. But in the system, the data stored in the cloud
is also encrypted using the Paillier encryption algorithm [26].
The Paillier algorithm has a unique feature of additive
homomorphic property shown in Eq. 6. The whole
communication in the proposed system is done in encrypted
form only.

Encrypt (P1 ⊕ P2) = Encrypt (P1) ⊕ Encrypt(P2) (6)

Figure 5. Elliptic curve

346
 A and B is represented as a geometric distance in Eq. 8 and
Algorithm - Generate public and private key
Public key =q minutiae angle difference in Eq. 9.
Private key =d
G(x,y) - base point
Begin Claimed
Identity
1. Select a random or pseudo-random integer i such that 1 ≤ d ≤ n−1.
2. Compute q = dG.
3. return (q, d).
End
Extract
Matcher Database
Features

Figure 6. Generation of keys

True or False

Algorithm - Signature generation

Input Required - Figure 9. Verification Phase

Private key d
Base point G(x,y)
n - order n of P (point on elliptic curve) A= (mA1,mA2,.......,mAp) (6)
Begin
1. Select k using random number generator
2. (x1, y1)= k * G(x,y) mod p where mAi = (xAi,yAi,θAi) and 1 ≤ i ≤ p
3. r = x1 mod n
4. if r=0 then
goto step 1 B= (mB1,mB2,.......,mBq) (7)
5. s=(k-1 (h(m) + d * r) mod n)
6. if s=0 then
goto step 1 where mBj = (xBj,yBj,θBj) and 1 ≤ j ≤ q
7. return(r,s)
End
2 2
𝐷𝑖𝑠𝑡𝑟 (𝑚𝐴𝑖 , 𝑚𝐵𝑗 = √(𝑥𝐴𝑖 − 𝑥𝐵𝑗 ) + (𝑦𝐴𝑖 − 𝑦𝐵𝑗 ) < 𝑟𝛼 (8)
Figure 7. Steps of signature generation
𝐷𝑖𝑠𝑡𝜃 (𝑚𝐴𝑖 , 𝑚𝐵𝑗 = min(|𝜃𝐴𝑖 − 𝜃𝐵𝑗 |, 3600 − |𝜃𝐴𝑖 −
Algorithm - Signature verification
𝜃𝐵𝑗 |) < 𝑟𝛼 (9)
Input Required -
Private key d
Signature(r,s) Algorithm -Calculate Similarity score
Base point G(x,y) T1 - stored image transformed minutiae
Message digest h(m) T2 - query image transformed minutiae
Begin T= 15 (Threshold for distance)
1. w = s-1 mod n TT=14 (Threshold for theta)
2. u1 = (h(m) * w) mod n Begin
3. u2 = (r * w) mod n Count1=size(T1,1)
4. (x2, y2) = (u1 × G(x, y) + u2 × Q(x, y)) mod n Count2=size(T2,1)
5. if x2 = r then n=0
6. print Successful For i=1 to Count1 do
7. else Found=0; j=1;
8. print Reject while (Found==0) and (j<=Count2) do
9. Endif dx=(T1(i,1)-T2(j,1));
End dy=(T1(i,2)-T2(j,2));
d=sqrt(dx2+dy2); //(Euclidean Distance between T1(i) & T2(i))
if d<T then
distheta=abs(T1(i,3)-T2(j,3))*180/pi;
Figure 8. Steps of signature verification distheta=min(distheta,360-distheta);
if distheta<TT then
n=n+1; (increase score)
3.2 Verification phase Found=1;
End if
End if
After enrollment of all the fingerprint of authorized users, j=j+1;
verification will be done each time a user wishes to access End While
End For
cloud data. The verification process is shown in Figure. 9 sm=sqrt(n2/(Count1*Count2)); //(Similarity Score)
Verification process is done by extracting minutiae points of a End

user, who wants to access the cloud system. After the

extraction of the features, the matching score also called a
Figure 10. Calculation of Similarity Score
similarity score is calculated for the query image with each
template existent in the database. This similarity score
describes the level of similarity between two fingerprints. The
4. RESULTS AND DISCUSSIONS
matching process Algorithm is shown in Figure. 10, which
compares both minutiae point sets viz. Input image I= m1,
4.1 Experiment setup
m2, .m3- mi) and template stored in database T= m1, m2, m3
- mj). The algorithm then returns a similarity score of T and I
The simulations of the proposed work are done in MATLAB
represented by S (I, T). Two minutiae points are called as
R2015a using a workstation with Intel coreTM i3 3.2 GHz
matched points if the calculated difference of position and
processor. The proposed approach has been applied on the
direction are less then acceptance distances. Let A and B are
dataset of 300 fingerprint grayscale images, captured from the
two images, where A is a sample image from the dataset and
employees of a University situated in north India. For the
B is the query image. A and B is a set of minutiae points
researcher's comfort, the collected images have been resized
represented by Eq. 6 and 7 respectively. The position between

347
to a square image of size 300×300. To measure the which any system incorrectly predicts the unauthorized
performance of the system, we have split the database into two biometric entity as a correct entity, while FNMR is the
parts- training and testing. For training of the system 80% of probability at which any system predicts the right entity as
the data has been used. For testing, 20% of data has been wrong. The equation of FMR and FNMR has been shown in
chosen from the dataset. Eq. 10 and 11 and the plots of FMR and FNMR of the sample
query image have been shown in Figure. 13. From Figure. 13,
4.2 Performance evaluation and comparison it can be clearly observed that the FMR and FNMR rate getting
better when repeated for 100 iterations. Equal Error Rate (EER)
The minutiae point algorithm is used for matching the is approximately 0.38 where FMR and FNMR value are equal.
fingerprint and finding the similarity score for the individual After applying the PCA algorithm the system produces
users. For generating image template original image is approximately 97% of accuracy, shown in Table 1 and the time
converted into masked, thin and then minutiae-points image is taken in the encryption of biometric features used for
generated Figure. 11 shows the masked, thinned and minutiae- recognition has been shown in Table 2.
point generated images along with four sample input images
chosen from the database. Figure. 12 displays the similarity
score after matching the minutia points of query and two (10)
template images taken from the cloud database. Three
verification metrics namely, False Matching Ratio (FMR),
False Non Matching Ratio (FNMR), and Recognition Rate (11)
(RR) have been determined. FMR determines a probability at

Figure 11. Mask, thinned and minutiae points for the four original input images

348
 Figure 12. Matching query and template image with similarity score 1 and Unmatched query and template image with
similarity

extracted features. The system has claimed 95% recognition

accuracy. In the study done by Kumar et al. [14], the authors
proposed a Biometric Face Recognition (BFR) system using
face detection. The authors have used the Eigenface detection
algorithm with the encryption using the elliptic curve. The
system claimed 96.89% recognition accuracy. Shakil et al.
[19], illustrated the cloud-based system for healthcare. The
system ensures the privacy and security of electronic medical
data. The system uses a signature-based biometric
authentication system. The authors have used the Back
Propagation Neural network (BPN) for the training of
signatures data. The system claimed sensitivity of 0.98 and a
specificity of 0.95. The study proposed by Balton et al. [27], a
cloud biometric system has been presented for authentication.
Figure 13. Plots of FMR and FNMR The system uses fingerprint and iris codes as biometric. The
system uses Minutiae points and Iris codes (MI) for feature
Table 1. Recognition accuracy of individual images extraction. The system claimed to consume the total time for
offline detection 3178 + 79.5/fused features. The summary
Sno. Image Encrypted Time Performance table of comparisons is shown in Table. 3, the accuracy plot is
No. Template (s) shown in Figure. 14, After done the comparison to all the
1 101 68,348,174,166 13 92.11 above mentioned studies here, the proposed system has
2 102 12,495,642,070 15 94.67 achieved an accuracy of 97%, which is showing a slight
improvement over others. But given the cost and advantages
3 109 24,579,125,485 18 95.57
over other biometrics systems as mentioned in the above
4 110 13,785,953,190 22 97
studies, the proposed system is better than the existing
approaches. Figure. 15, shows the recognition rate comparison
Table 2. Computation time for verification of individual plot between the proposed approach and other works. The
images figure shows clearly that the recognition rate of proposed work
is far better than other stated approaches in terms of time to
Sno. User Image Image size (pixels) Time Taken (s) recognize query fingerprint.
1 101 60 x 60 180
2 102 100 x 100 220
3 109 200 x 200 275
4 110 250 x 250 297

To validate the performance of the proposed system, we

have compared our approach with some existing approaches
given by various researchers such as Haghighat et al. [10],
Kumar et al. [14], Shakil et al. [19], and Balton et al. [27] etc.
In the research study proposed by Haghighat et al. [10], a cloud
system based biometric recognition for individual user
authentication has been presented. The system adds the details
of users to their biometric and store it after encryption. The
facial images are used as biometric. The system also uses the
Generalized Local Discriminant Analysis (GLDA) to classify Figure 14. Plot of performance comparison

349
 Figure 15. Plots of recognition rate comparison

Table 3. Comparative analysis of existing biometric systems

Sno. Study Biometric Approach Used Performance

1 GLDA Facial images GLDA Recognition accuracy -95 %
Recognition accuracy-
2 BFR Face images Eigen face, encryption
96.89 %
3 BPN Signatures Sensitivity 0.98, specificity
Backpropagation neural network
0.95.
4 MI Fingerprint and iris codes Minutiae points and iris codes 3178 +79.5 offline detection
time 89 +
149.2 online detection time
5 Proposed Fingerprints and encryption with digital signature Minutiae points, PCA , Hybrid two Recognition accuracy 97 %
Study and homomorphic encryption layer encryption

5. CONCLUSION AND FUTURE SCOPE authentication in the cloud with CDVS descriptors.
Pattern Recognition Letters, 113: 67-74.
In this paper, a secure and privacy-preserving cloud system https://doi.org/10.1016/j.patrec.2017.03.024
has been proposed, which is based on a hybrid biometric [2] Jain, A.K., Ross, A.A., Nandakumar, K. (2011).
recognition system and elliptic curve cryptography. The Introduction to biometrics. Springer Science & Business
system identifies cloud users according to their encrypted Media. https://doi.org/10.1007/978-0-387-77326-1
fingerprint templates stored in the encrypted domain. For [3] Ratnam, S., Gupta, M., Singh, D.A.S. Thirunavukkarasu,
feature extraction, a minutiae point detection algorithm is used K. (2016). A survey on biometric security technologies
which uses two features ridge ending and ridge bifurcations. from cloud computing perspective. International Journal
The query image can be recognized according to the proposed of Scientific and Technology Research, 4(8): 22–24.
algorithm which generates a similarity score in terms of FMR [4] Jain, A.K., Pankanti, S., Prabhakar, S., Hong, L., Ross,
and FNMR which lies between 0 to 1. To improvise the A., Wayman, J.L. (2004, August). Biometrics: A grand
recognition accuracy by reducing the noise PCA approach has challenge. Proceedings of the 17th International
been applied to the proposed system. After experimental Conference on Pattern Recognition, Cambridge, UK, pp.
evaluation of the proposed scheme, it has been found that the 935-942. https://doi.org/10.1109/ICPR.2004.1334413
system recognition accuracy is approximately 97% which is [5] Jain, A.K., Ross, A., Pankanti, S. (2006). Biometrics: A
quite better than state-of-art recent approaches. The main tool for information security. IEEE transactions on
shortcoming of the system is the storage requirement. As the Information Forensics and Security, 1(2): 125-143.
system goes in real time, the database size requirement gets https://doi.org/10.1109/TIFS.2006.873653
increased significantly because of the large size of images in [6] Jain, P., Rane, D., Patidar, S. (2011). A survey and
comparison to traditional authentication data. Further, a small analysis of cloud model-based security for computing
dataset has been chosen for testing purpose which can be taken secure cloud bursting and aggregation in renal
into consideration in the future. Also, a combination of one or environment. In 2011 World Congress on Information
more traditional features or biometric parameters like and Communication Technologies, IEEE, Mumbai, India,
passwords, retina scan, signature, etc. can be added to make pp. 456-461.
the system more robust and secure. https://doi.org/10.1109/WICT.2011.6141288
[7] Gupta, B., Agrawal, D.P., Yamaguchi, S. (2016).
Handbook of Research on Modern Cryptographic
REFERENCES Solutions for Computer and Cyber Security. IGI Global.
https://doi.org/10.4018/978-1-5225-0105-3
[1] Fiandrotti, A., Mattelliano, M., Baccaglini, E., Vergori, [8] Bhattasali, T., Saeed, K., Chaki, N., Chaki, R. (2015). A
P. (2018). CDVSec: Privacy-preserving biometrical user survey of security and privacy issues for biometrics

350
 based remote authentication in cloud. In IFIP https://doi.org/10.1016/j.future.2016.10.004
International Conference on Computer Information [18] Al-Hamami, A.H., AL-Juneidi, J.Y. (2015). Secure
Systems and Industrial Management, Springer, Berlin, mobile cloud computing based-on fingerprint. World of
Heidelberg, pp. 112-121. https://doi.org/10.1007/978-3- Computer Science & Information Technology Journal,
662-45237-0_12 5(2): 23-27.
[9] Naveed, G., Batool, R. (2015). Biometric authentication [19] Shakil, K.A., Zareen, F.J., Alam, M., Jabin, S. (2017).
in cloud computing. Journal of Biometrics & BAMHealthCloud: A biometric authentication and data
Biostatistics, 6(5): 1. https://doi.org/10.4172/2155- management system for healthcare data in cloud. Journal
6180.1000258 of King Saud University-Computer and Information
[10] Haghighat, M., Zonouz, S., Abdel-Mottaleb, M. (2015). Sciences. https://doi.org/10.1016/j.jksuci.2017.07.001
CloudID: Trustworthy cloud-based and cross-enterprise [20] Maltoni, D. (2003). A tutorial on fingerprint recognition,
biometric identification. Expert Systems with advanced studies in biometrics. Summer School on
Applications, 42(21): 7905-7916. Biometrics, Alghero, Italy.
https://doi.org/10.1016/j.eswa.2015.06.025 https://doi.org/10.1007/11493648_3
[11] Hahn, C., Hur, J. (2016). Efficient and privacy- [21] Wieclaw, L. (2009). A minutiae-based matching
preserving biometric identification in cloud. ICT Express, algorithms in fingerprint recognition systems. Journal of
2(3): 135-139. https://doi.org/10.1016/j.icte.2016.08.006 Medical Informatics & Technologies, 13.
[12] Bala, Y., Malik, A. (2018). Biometric inspired [22] Rutovitz, D. (1966). Pattern recognition. Proceedings of
homomorphic encryption algorithm for secured cloud Journal in Royal Statistical Society, vol. 129.
computing. In Nature Inspired Computing, Springer, https://doi.org/10.2307/2982255
Singapore, pp. 13-21. https://doi.org/10.1007/978-981- [23] Wang, Y.X., Ao, X.Y., Du, Y.F., Li, Y.P. (2006). A
10-6747-1_2 fingerprint recognition algorithm based on principal
[13] Pan, S., Yan, S., Zhu, W.T. (2016, July). Security component analysis. In TENCON 2006-2006 IEEE
analysis on privacy-preserving cloud aided biometric Region 10 Conference, Hong Kong, China, pp. 1-4.
identification schemes. In Australasian Conference on https://doi.org/10.1109/TENCON.2006.344032
Information Security and Privacy, Springer, Cham, pp. [24] Martinez, V.G., Encinas, L.H., Ávila, C.S. (2010). A
446-453. https://doi.org/10.1007/978-3-319-40367-0_29 survey of the elliptic curve integrated encryption scheme.
[14] Kumar, S., Singh, S.K., Singh, A.K., Tiwari, S., Singh, Ratio, 80(1024): 160-223.
R.S. (2018). Privacy preserving security using biometrics [25] Shankar, T.N., Sahoo, G., Niranjan, S. (2012). Using the
in cloud computing. Multimedia Tools and Applications, digital signature of a fingerprint by an elliptic curve
77(9): 11017-11039. https://doi.org/10.1007/s11042- cryptosystem for enhanced authentication. Information
017-4966-5 Security Journal: A Global Perspective, 21(5): 243-255.
[15] Lee, P. (2017). Prints charming: how fingerprints are https://doi.org/10.1080/19393555.2012.694978
trailblazing mainstream biometrics. Biometric [26] Paillier, P. (1999, May). Public-key cryptosystems based
Technology Today, 2017(4): 8-11. on composite degree residuosity classes. In International
https://doi.org/10.1016/S0969-4765(17)30074-7 Conference on the Theory and Applications of
[16] Zhang, C., Zhu, L., Xu, C. (2017). PTBI: An efficient Cryptographic Techniques, Berlin, Heidelberg, pp. 223-
privacy-preserving biometric identification based on 238. https://doi.org/10.1007/3-540-48910-X_16
perturbed term in the cloud. Information Sciences, 409: [27] Blanton, M., Gasti, P. (2011). Secure and efficient
56-67. https://doi.org/10.1016/j.ins.2017.05.006 protocols for iris and fingerprint identification. In
[17] Kumari, S., Li, X., Wu, F., Das, A.K., Choo, K.K.R., European Symposium on Research in Computer Security,
Shen, J. (2017). Design of a provably secure biometrics- Springer, Berlin, Heidelberg, pp. 190-209.
based multi-cloud-server authentication scheme. Future https://doi.org/10.1007/978-3-642-23822-2_11
Generation Computer Systems, 68: 320-330.

351