Journal of Ambient Intelligence and Humanized Computing (2021) 12:6141–6149

https://doi.org/10.1007/s12652-020-02184-8

ORIGINAL RESEARCH

A multimodal biometric authentication scheme based on feature

fusion for improving security in cloud environment
Teena Joseph1 · S. A. Kalaiselvan2 · S. U. Aswathy3 · R. Radhakrishnan4 · A. R. Shamna5

LE
Received: 5 March 2020 / Accepted: 5 June 2020 / Published online: 23 June 2020
© Springer-Verlag GmbH Germany, part of Springer Nature 2020

Abstract

C
In recent days, due to the advent of advanced technologies such as cloud computing, accessing data can be done anywhere at
any time. Meanwhile, ensuring the data security is highly significant. Authentication plays a major role in preserving security
via different access control mechanisms. As a recent trend, the biological information of the individual user is considered

TI
as verification scheme for the authentication process. Traits such as fingerprint, iris, ear or palm print are widely used to
develop the authentication systems from its patterns. But, to increase the complexity of the user authentication and to ensure
high security, more than a trait is combined together. In this paper, a multimodal authentication system is proposed by fus-

R
ing the feature points of fingerprint, iris and palm print traits. Each trait has undergone the following procedures of image
processing techniques such as pre-processing, normalization and feature extraction. From the extracted features, a unique
secret key is generated by fusing the traits in two stages. False Acceptance Rate (FAR) and False Rejection Rate (FRR)
A
metrics are used to measure the robustness of the system. This performance of the model is evaluated using three standard
symmetric cryptographic algorithms such as AES, DES and Blowfish. This proposed model provides better security and
access control over data in cloud environment.
D
Keywords Authentication · Cloud computing · Cryptography · Hashing · Multimodal biometric system · Symmetric key
TE

encryption

1 Introduction
C

* Teena Joseph Cloud computing is an advanced technology, which deliver

[email protected] services without direct management of user, based on the
A

S. A. Kalaiselvan demand of resources (Kushida et al. 2015). It is highly scal-

[email protected] able, robust and provides access to the data anywhere at any
S. U. Aswathy time. It supports performing complex, high-scale operations
R

[email protected] over cloud environment. The key advantage of this technol-

R. Radhakrishnan ogy is in ensuring better resource management, access con-
ET

[email protected] trol and security. The service provided by cloud is expanding

A. R. Shamna in different form such as Platform as a Service (PaaS), Infra-
[email protected] structure as a Service (IaaS), Software as a Service (SaaS)
1 etc. (Armbrust et al. 2010). So, day by day, the number of
Department of Computer Science and Engineering, MBC
users consuming the services of cloud is increased.The data
R

CET, Peermade, India

2 can be stored in different cloud services where it can be
St Martin’s Engineering College, Secunderabad, Telegana,
India accessed remotely by the user, whenever it is needed. But
3 the major concern is in maintaining data security (Chang
Department of Computer Science and Engineering, Jyothi
Engineering College, Thrissur, India and Ramachandran 2016; Stallings 2017). Since the data
4 is in remote server, it is prone to any malicious attacks and
Department of Civil Engineering, Mangalam College
of Engineering, Kottayam, India sometimes can be compromised. So, developing a highly
5 secure data authentication and access control mechanism is
Musaliar College of Engineering, Chirayinkeezhu, India

13
Vol.:(0123456789)
6142 T. Joseph et al.

inevitable. In traditional service providing models, the secu- its importance. Section 5 concludes the work by highlight-
rity mechanisms are built on top of standard cryptographic ing the findings.
algorithms with highly complex mathematical calculations
(Panchal et al. 2017; Ross and Jain 2004). But modern day
computing technologies are highly sophisticated, which can 2 Background study
be significant enough to break the security of most of the
systems built before a decade. In order to enhance the secu- Multimodal biometric authentication system is an integral
rity in cloud environment, various schemes are introduced part of multiple domains such as cryptography, image pro-
(Latha and Sheela 2019). cessing and computer networks. A multimodal biometric
Biometrics refers to physical measurement and its related system based authentication system is developed by fusing

LE
calculations of human body. These metrics can be directly fingerprint and palm print. Histogram equalization is made
related to the characteristics, otherwise modalities of the to normalize the pixel intensity. For feature extraction, PCA
human. Also, these metrics are used in most of the authen- is used and fusion of features is carried out by Gabor wavelet
tication schemes via access control systems (Faryadi 2013; method. This model is developed to strengthen up the secu-
Asthana et al. 2014). Some traits are said to be fingerprint, rity on user authentication process (Preneel 1994; Delac and

C
iris, palm print, retina, DNA, voice, gait etc. The traits of Grgic 2004; Gayathri and Ramamoorthy). A hash key based
every individual are collected, processed, stored in a data- cryptographic algorithm is proposed by combining the fea-

TI
base and this process is called as enrollment. Then during tures extracted from multimodal biometrics to enhance cloud
verification process, the user is authenticated via an access security. Fingerprint, iris and face modalities are incorpo-
control scheme such as fingerprint authentication. The rated in this system. Features of independent modalities are
importance of biometrics lies in its uniqueness in the pat- extracted using various image processing techniques. To

R
tern of every individual. Uni-modal biometric systems are optimize the extracted features, a swarm intelligence tech-
fairly outdated, have many drawbacks and used only in some nique called Artificial Fish Swarm (AFS) algorithm is used.
scenarios where the security is not a big concern (Thakur
and Kumar 2011; Jeyakumar et al. 2018). Fusing more than
a trait is called a Multimodal biometric system. The robust-
A
The data to be stored in cloud is encrypted by AES algo-
rithm with the generated hash key (Oravec 2014).
A multimodal biometric-based cloud security enhance-
D
ness of the system is high with multimodal authentication ment technique is developed by fusing three modalities. The
process. fused vectors are binarized, then XOR is performed between
Cryptography otherwise cryptology is a concept of con- the transformed feature vectors. The final resultant key is
TE

verting a normal plain information from human-readable used to encrypt the data before storing in the cloud using
form to an unreadable format (Kamalakannan et al. 2015; AES (Mansour et al. 2015). A robust authentication system
Ranjith et al. 2016). This process is called as encryption. for user verification is framed from face and iris modalities.
The reverse mechanism of the same process is called as Machine Learning algorithms are employed to detect the
C

decryption. Some common types of this cryptographic algo- feature vectors from the image. Multilayered Perceptron and
rithm are symmetric key cryptography (SKC), asymmetric Self Organized Map neural networks are used in this process
key cryptography (AKC), and hash functions etc. (Asthana (Khatri and Vadi 2017; Masala et al. 2018).
A

et al. 2014). In SKC, the sender and receiver have a com- Multifactor authentication system on cloud environment
mon secret key to encrypt and decrypt the messages. Anyone is deployed with multimodal biometric scheme (MFA-
R

who is having the secret key can read the message (Thakur MB) (Ahmad et al. 2017). Password-based user verifica-
and Kumar 2011). But in AKC, two keys such as public tion makes the data security on cloud as incomplete. This
and private keys are involved in crypto process (Kamala- system enhances the security with multiple levels of user
ET

kannan et al. 2015). The public key is shared to everyone, authentication such as password and multimodal biometric
whereas the private key is confidential. If both the key are data (Sarier 2017). In a similar way, a two-step verification
valid, then the message can be decrypted. Hashing is used scheme is proposed to improve cloud security on mobile
only when the message needs to be converted to a collection communication. Iris and fingerprint modalities are used in
R

of characters or numbers, which cannot be transformed back this system for authentication of the user (Nair et al. 2017;
to its original form. In simple terms, hashing is a “one-way- Mansour et al. 2016).
encryption”, where the message cannot be decrypted again A biometric-based cloud authentication system is
(Ushapreethi et al. 2017; Inani et al. 2018). depicted to provide access control and ensuring data secu-
This paper is organized as follows: Sect. 2 discusses the rity using various schemes for different cloud architectures
concepts related to this work. In Sect. 3, the workflow of (Vidya and Chandra 2018). To preserve privacy in cloud,
the proposed model is clearly stated in detail. The result a multimodal biometric system is introduced to resist hill
obtained from this work is depicted in Sect. 4 and discusses climbing attacks. Two-way encryption is performed in this

13
A multimodal biometric authentication scheme based on feature fusion for improving security… 6143

work. In step-1, uni-modal or multimodal biometric encryp- are performed to extract feature region from the input fin-
tion is performed for first level of authentication. The suc- gerprint data.
cessor step follows the process of calculating the overlapping
information on the data using Euclidean distance measure as (i) Image normalization and segmentation: The gray-
next level (Abed et al. 2018). level image is undergone some adjustment in the
A decision level fusion mechanism for multi-biometric range of value to standardize the intensity of the
cryptosystem is proposed to perform secure file upload in image.
cloud. This system provides high security over authenti- (ii) Orientation estimation: Initially the gradient is cal-
cation and integrity on access control to handle the data culated for each pixel in the image. Further, the local
(Kundu and Sarker 2016). Context-aware models to secure orientation is calculated by finding the variation axis

LE
cloud empowered systems are deployed using multimodal in the gradients of the image. The orientation field in
biometric methods. Three level of process is done in this sys- the image is then smoothened by Gaussian low pass
tem, first is to integrate class association rule into MFA-MB. filter method. Figure 1 shows themultimodal biom-
Then a new metric is developed to evaluate the user experi- etric authentication system framework.
ence on this system and finally to enhance the authentication (iii) Morphological processing: Before extracting the

C
on PaaS and SaaS provided by cloud (Arrawatia et al. 2017). minutiae points from the fingerprint, image thinning
An entropy-based Local Binary Pattern technique is process is done. It highlights the ridges and maps the

TI
developed in this work to extract the features from multi- feature points. The outcome of this step is the genera-
modal traits. The extracted features are highly robust to ver- tion of a skeletal pattern of ridges.
ify the users with low FAR and high FRR. This mechanism (iv) Extracting region of interest (ROI): The feature
is embedded in cloud scenario to enhance data security (Jia points are extracted from the skeletal image using

R
et al. 2017). The feasibility factor for generating key using two morphological operations such as ERODE and
different modalities is investigated in this work. Fingerprint, OPEN. Figure 2 shows the fingerprint feature point
keystrokes and face modalities are used to test the model
and achieved a fair result as an outcome. A transparent key
generation technique is introduced to frame a key of length
A extraction method.

4.1.2 Iris
D
256 bits (Aithal 2018).
Iris is a highly significant part of human eye. The structure
inside the human eye varies between everyone. This is one
TE

3 Materials and methods of the robust ways to authenticate a user in highly secure
processes (Zhang and Parhi 2002). Many approaches are
The database used in this work is framed by collecting the successful in extracting region of interest from iris. In this
biometric image samples from 100 individuals for all three system, some methods are followed, which are given below.
C

traits. The images are cropped into equal m * n size to main- Figure 3 shows theiris feature point extraction method.
tain consistency among the pixels. Further processing of
image is stated in depth while discussing each modality in (i) Binarization: As an initial processing of iris image,
A

upcoming sections. the data is binarized using Otsu thresholding tech-

nique.
R

(ii) The edges of the image curve are traced out by using
4 Proposed multimodal biometric canny edge detection method.
authentication system framework (iii) Boundaries of edge detected image are identified
ET

using Hough transform.

4.1 Extraction of patterns from different traits (iv) Features from the iris part is extracted using Gabor
filtering technique
4.1.1 Fingerprint
R

4.1.3 Palm print
The basic form of biometric trait used for authentication
is fingerprint. Mostly the pattern of fingerprint is unique Tracking the palm region from the hands is tedious due to
to every individual. So, these features are considered for the noise, intra-class variation and thin ridges (Schneier
verification process. The fingerprint image taken from the 1993). But this method ensures the integrity of the authen-
sensor might have noise and sometimes occlusion (Schaefer tication system due to its uniqueness. This trait can be pro-
1996). To eliminate these issues, some preprocessing tech- cessed in the way listed out in upcoming steps. Figure 4
niques are applied on the image data. The following steps shows thepalm print feature point extraction method.

13
6144 T. Joseph et al.

Fig. 1  Multimodal biometric

authentication system frame-
work

LE
C
TI
R
A
D
Fig. 2  Fingerprint feature point
extraction method
TE

Fig. 3  Iris feature point extrac-

tion method
C
A

Fig. 4  Palm print feature point

extraction method
R
ET

(i) Image taken from the sensor is initially smoothened 4.2 Key generation using XOR and hashing
using Gaussian method techniques
(ii) Then, the smoothened image is binarized using Otsu
thresholding method. 4.2.1 Binarization
R

(iii) To detect the regions of palm, Sobel edge detection

technique is applied on binarized image. After the feature points are extracted from all the traits,
(iv) Using 2D—Gabor Wavelet technique, the feature it is converted further into binary pattern. Based on the
points from the palm region is extracted. threshold value for each pixel in the identified features, it
will be either replaced by either 0 or 1. The final feature
vector will be transformed as a sequence of binary values.

13
A multimodal biometric authentication scheme based on feature fusion for improving security… 6145

4.2.2 XOR 4.2.3 MD‑5 hashing

In this module, three binary value calculated from each MD-5, otherwise Message-Digest algorithm is a most
trait will be fused together in two stages. In stage 1, binary commonly used hashing technique. It produces a key with
equivalent of fingerprint and iris feature vector is combined length of 265 bits. It is mainly used to ensure the integrity
to form a new vector using XOR operation. Then, the newly of authentication process such as password encryption.
calculated value is again undergone XOR operation with The resultant XOR feature vector is finally converted into
the binary equivalent of third trait, i.e. palm print. A unique a hash string of length 128 bits using MD-5. This hash key
binary key will be the resultant of this process. Figure 5 is the final secret key used for data encryption and decryp-
shows the secret key generation process from multimodal tion process in cloud environment.

LE
biometric system.

C
TI
R
A
D
TE
C
A
R
ET
R

Fig. 5  Secret key generation process from multimodal biometric system

13
6146 T. Joseph et al.

LE
C
TI
R
A
D
TE
C
A

5 Performance of cryptographic algorithms 5.2 Advanced encryption standard

on generated biometric key
R

AES is a well-known symmetric key cryptographic algo-

5.1 Data encryption standard rithm, widely accepted to develop any security applica-
tion. It is six times faster when compared with triple DES
ET

DES algorithm is developed by National Institute of Stand- algorithm. The drawback of DES is in its minimal key size.
ards and Technology (Kundu and Sarker 2016). It is a sym- In AES, 128-bit key is used to encrypt the data (Karthik
metric encryption algorithm (Sajay et al. 2019), widely et al. 2018). Since the size of key is high, it makes the
used before AES. Feistel cipher is the base part of DES system more complex to crack it. Also the size of the
R

implementation. In DES, Feistel structure with 16 round key can be extended up to 192 and 256 bit. The process
patterns is followed. Size of the block is 64 bit. But the involved in AES is iterative. Substitution and Permutation
optimal key length is 56. Because from 64-bit key, 8 bit are the major part in AES working process. AES performs
are not used by DES during encryption process. The key its calculations in bytes rather than bits. So, it is easy to
components of DES are round function, key schedule and frame the matrix. The number of rounds in AES varies
initial/final permutation. A main challenge in DES is, it depends on the size of the key. For the key with size 128
is prone to exhaustive key search, which is performed to bits, the number of rounds is 10. Similarly for 192-bit key,
make a cryptanalytic attack on the data. 12 rounds and 256-bit key, it is 14 rounds. Till now, no

13
A multimodal biometric authentication scheme based on feature fusion for improving security… 6147

Table 1  False Acceptance Rate

Image count Fingerprint Iris Palm print Fused features

1–20 0.39 0.36 0.36 0.11

21–40 0.42 0.29 0.32 0.14
41–60 0.36 0.35 0.35 0.13
61–80 0.33 0.26 0.39 0.16
81–100 0.38 0.41 0.31 0.21
Average 0.376 0.334 0.346 0.15

LE
Table 2  False Rejection Rate
Image count Fingerprint Iris Palm print Fused features

1–20 92.26 94.60 92.62 97.86

C
21–40 91.05 96.85 91.56 94.58 Fig. 6  Comparison of the performance of different modalities using
41–60 88.92 93.28 89.95 96.25 FAR

TI
61–80 90.23 91.35 91.96 91.36
81–100 89.98 93.67 90.30 92.65
Average 90.488 93.95 91.278 94.54

R
Table 3  Execution time of three algorithms in CPU (in milliseconds)
File size/algorithm DES
(key size = 56
bits)
AES Blowfish
(key size = 128 (key
bits) size = 128
A
D
bits)

512 KB 22 26 23
TE

1 MB 25 29 26
3 MB 31 38 35
5 MB 37 44 40
C

successful attacks can be performed on AES, which exhib-

its the strength and security of the algorithm. Fig. 7  Comparison of the performance of different modalities using
FRR
A

5.3 Blowfish
R

In this proposed work, initially the features from three

Blowfish is a symmetric key encryption algorithm, which different traits are extracted by some unique image pro-
is used as a replacement of DES algorithm. The length cessing strategies. In next step, the feature points are con-
ET

of the key varies from the range of 32 bits to 446 bits. verted to its binary equivalent to perform XOR operation
The block size of this algorithm is 64 bit. It follows Feis- between the traits. XOR is performed with fingerprint
tel cipher with 16 rounds (Karthik and Sudha 2018). It is and iris data. The resultant binary vector is then under-
one of the fastest block ciphers in symmetric algorithms. gone XOR with binarized palm print feature vector. The
R

The block size of this algorithm is 64 bit, which is half obtained key after these processes are then converted to
in size when compared with AES block size. So, in some has of strings and numbers using MD-5 hashing. This gen-
cases, it is vulnerable to some cryptanalytic attacks. A erated secret key is then used to perform data encryption
cryptographic algorithm with similar mechanism is devel- and decryption. The size of the binary key generated by
oped called as Twofish, which is a successor of Blowfish MD-5 is usually 256 bits. The first 56 bits are used in DES
algorithm. and first 128 bits are used in both AES and Blowfish for
data encryption.

13
6148 T. Joseph et al.

This work focuses on preserving high data security on

cloud environment. So, from the three algorithms, AES
ensures better security due to its complex computational
steps. Next to that, Blowfish is considered and DES reserves
least position.

7 Conclusion

Authentication is an important factor on ensuring security

LE
for various applications. Cloud computing is an internet-
based model for providing service to various end users
Fig. 8  Benchmarking the execution time for different file size using related to information technology. It provides high flexibility
three algorithms for the users, so the usability of cloud services is increasing
gradually. Also, it makes the concern about data security

C
to some extent. Multimodal biometric system enhances the
6 Results and discussion robustness of the authentication mechanism because of its

TI
inherent unique biological patterns. It accurately discrimi-
The proposed model is evaluated using two standard metrics nates the individuals based on the captured pattern from
such as False Acceptance Rate (FAR) and False Rejection Rate their traits. Moreover, this concept can be applied in various
(FRR) calculated from the Eqs. 1 and 2. In Tables 1 and 2, applications to improve the robustness of the system such

R
FAR and FRR is calculated for every individual trait also after as securing human genetic code and health information for
fusing the features respectively. From the results, it is identi- future reference using Electronic Health Record (EHR) man-
fied that FAR is very less when the features are fused together
based on the mechanism given in the proposed model. Iris has
less FAR next to fused features. Palm print and fingerprint
A
agement, digital ledger management, etc.
In this work, a multimodal biometric authentication
mechanism is proposed to make data in cloud environment
D
bags the next consecutive positions. Meanwhile, the same is more secure. A secret key is generated by fusing the features
reflected in FRR, where the proposed fusion methodology out- extracted from fingerprint, iris and palm print in multiple
performed uni-modal techniques. Similar to FAR, iris, palm stages and finally converted into hash of strings and num-
TE

print and fingerprint shows better results next to fused feature bers using MD-5 hashing algorithm. The data to be secured
in consecutive order. Table 3 shows the execution time of three is then encrypted by the secret key with three symmetric
algorithms in CPU. key encryption algorithms DES, AES and Blowfish. Among
them DES takes less execution time, but AES has better per-
FP
C

FAR = (1) formance when compared with other two algorithms based
FP + TN on the strength of encryption process. This model proved
its robustness in data security due to the fusion of human
A

FRR =
FN
(2) modalities as a part of framing the security mechanism.
FN + TP
R

where, TP, FN, FP, TN represents True Positive, False

Negative, False Positive and True Negative respectively. References
ET

Figures 6 and 7 represents the values in graph plot to

visually evaluate the results. A comparison is made between Abed L, Clarke N, Ghita B, Alruban A (2018) Securing cloud storage
three symmetric key encryption algorithms to measure the by transparent biometric cryptography. In: International Confer-
efficiency and performance of each on the proposed secret ence on Security for Information Technology and Communica-
tions, pp 97–108.
key strategy based on execution time under different file Ahmad A, Paul A, Khan M, Jabbar S, Rathore MMU, Chilamkurti N,
R

sizes. Among them, DES consumes less time due to the Min-Allah N (2017) Energy efficient hierarchical resource man-
small size of its key and block size, where AES takes a lot agement for mobile cloud computing. IEEE Trans Sustain Comput
because of larger key size. Blowfish algorithm maintains sta- 2(2):100–112
Aithal PS (2018) A study on fingerprint hash code generation based on
ble performance in between both of it. But when comparing MD5 algorithm and freeman chain code. Int J Comput Res Dev
the security and integrity, AES tops the list. Figure 8 shows (IJCRD) 3(1):13–22
thebenchmarking the execution time for different file size Armbrust M, Fox A, Griffith R, Joseph AD, Katz R, Konwinski A,
using three algorithms. Zaharia M (2010) A view of cloud computing. Commun ACM
53(4):50–58

13
A multimodal biometric authentication scheme based on feature fusion for improving security… 6149

Arrawatia S, Mitra P, Kishore B (2017) Critical literature survey on Masala GL, Ruiu P, Grosso E (2018) Biometric authentication and data
iris biometric recognition. Int J Sci Res Sci Technol 3(6):600–605 security in cloud computing. Comput Netw Secur Essen 337–353.
Asthana R, Verma N, Ratan R (2014) Generation of Boolean functions Nair VS, Reshmypriya GN, Rubeena MM, Fasila KA (2017) Multibio-
using Genetic Algorithm for cryptographic applications. IEEE Int metric cryptosystem based on decision level fusion for file upload-
Adv Comput Conf (IACC) 1361–1366. ing in cloud. In: International Conference on Recent Advances
Chang V, Ramachandran M (2016) Towards achieving data security in Electronics and Communication Technology (ICRAECT), pp
with the cloud computing adoption framework. IEEE Trans Serv 29–32.
Comput 9(1):138–151 Oravec M (2014) Feature extraction and classification by machine
Delac K, Grgic M (2004) A survey of biometric recognition meth- learning methods for biometric recognition of face and iris. In:
ods. In: 46th International Symposium Electronics in Marine, pp International Symposium ELMAR (ELMAR), pp 1–4.
16–18. Panchal G, Samanta D, Barman S (2017) Biometric-based cryptogra-
Faryadi DQ (2013) Does data security matter? The case for cryptogra- phy for digital content protection without any key storage. Mul-
timed Tools Appl 1–22.

LE
phy. Int Conf Comput Sci Comput Math (ICCSCM) 1–6.
Gayathri R, Ramamoorthy P (2012) A fingerprint and palmprint rec- Preneel B (1994) Cryptographic hash functions. Eur Trans Telecom-
ognition approach based on multiple feature extraction. Eur J Sci mun 5(4):431–448
Res 76(4):514–526 Ranjith D, Balajee J, Kumar C (2016) In premises of cloud computing
Inani A, Singh M, Saxena R (2018) A secure mobile cloud comput- and models. Int J Pharm Technol 8(3):4685–4695
ing framework based on data classification using asymmetric key Ross A, Jain AK (2004) Multimodal biometrics: an overview. Signal

C
cryptography, pp 831–838. Process Conf 1221–1224.
Jeyakumar B, Durai MS, Lopez D (2018) Case studies in amalgamation Sajay KR, Babu SS, Vijayalakshmi Y (2019) Enhancing the security
of deep learning and big data. In: IGI global HCI challenges and of cloud data using hybrid encryption algorithm. J Ambient Intell

TI
privacy preservation in big data security, pp. 159–174. Humaniz Comput 1–10.
Jia W, Zhang B, Lu J, Zhu Y, Zhao Y, Zuo W, Ling H (2017) Palmprint Sarier ND (2017) Privacy preserving multimodal biometric authentica-
recognition based on complete direction representation. IEEE tion in the cloud. In: International Conference on Green, Perva-
Trans Image Process 26(9):4483–4498 sive, and Cloud Computing, pp 90–104.
Kamalakannan G, Balajee J, SrinivasaRaghavan S (2015) Superior Schaefer EF (1996) A simplified data encryption standard algorithm.

R
content-based video retrieval system according to query image. Cryptologia 20(1):77–84
Int J Appl Eng Res 10(3):7951–7957 Schneier B (1993) Description of a new variable-length key, 64-bit
Karthik S, Sudha MA (2018) Survey on machine learning approaches block cipher (Blowfish). In: International workshop on fast soft-
in gene expression classification in modelling computational
diagnostic system for complex diseases. Int J Eng Adv Technol
8(2):10–15
Karthik S, Perumal RS, Mouli PC (2018) Breast cancer classification
A ware encryption, pp 191–204.
Stallings W (2017) Cryptography and network security: principles and
practice. Pearson, Upper Saddle River
Thakur J, Kumar N (2011) DES, AES and Blowfish: symmetric key
D
using deep neural networks. In: Knowledge computing and its cryptography algorithms simulation based performance analysis.
applications, pp 227–241. Int J Emerg Technol Adv Eng 1(2):6–12
Khatri SK, Vadi VR (2017) Biometrie based authentication and access Ushapreethi P, JeyakumarBalaKrishnan BP (2017) Action recognition
TE

control techniques to secure mobile cloud computing. Int Conf in video surveillance using hipi and map reducing model. Int J
Telecommun Netw (TEL-NET) 1–7. Mech Eng Technol 8(11):368–375
Kundu S, Sarker G (2016) A new RBFN with modified optimal cluster- Vidya BS, Chandra E (2018) Entropy based Local Binary Pattern
ing algorithm for clear and occluded fingerprint identification. In: (ELBP) feature extraction technique of multimodal biomet-
International Conference Control, Instrumentation, Energy and rics as defence mechanism for cloud storage. Alexandria Eng J
Communication, pp 125–129. 58(1):103–114
C

Kushida KE, Murray J, Zysman J (2015) Cloud computing: from scar- Zhang X, Parhi KK (2002) Implementation approaches for the
city to abundance. J Ind Compet Trade 15(1):5–19 advanced encryption standard algorithm. IEEE Circuits Syst Mag
Latha K, Sheela T (2019) Block based data security and data distri- 2(4):24–46
A

bution on multi cloud environment. J Ambient Intell Humaniz

Comput 1–7. Publisher’s Note Springer Nature remains neutral with regard to
Mansour A, Sadik M, Sabir E (2015) Multi-factor authentication based jurisdictional claims in published maps and institutional affiliations.
R

on multimodal biometrics (MFA-MB) for cloud computing. IEEE/

ACS Int Conf Comput Syst Appl (AICCSA) 1–4.
Mansour A, Sadik M, Sabir E, Azmi M (2016) A context-aware mul-
ET

timodal biometric authentication for cloud-empowered systems.

In: International Conference Wireless Networks and Mobile Com-
munications (WINCOM), pp 278–285.
R

13