BIOMETRIC

Manpreet kaur (Research scholar) UIET, Panjab university, Chandigarh.

Dr. Amandeep Verma (Associate professor, UIET Panjab university Chandigarh)
Dr. Puneetjai kaur (Associate professor, UIET Panjab university Chandigarh)
Abstract: Security of any digital data is essential task. With the advancement of technology,
the methods for security are changed from time to time to patch the security breaches The Bio
information of human are combined with the technology for development of highly secure
systems known as biometric systems, which are used in different sectors to protect the digital
data. Various techniques of biometric related to physiological characteristics of humans are
developed which are used as contribution of secure system building. In this study, we are also
discussing the origin of biometric technology which is referred as first and second generation
of biometric systems. The system follows the four important basic steps for building block of
protection against the cyber-attacks on digital information. The complete process of biometric
Technology, characteristics and various attacks are discussed in detail. According to
circumstances various vulnerabilities are defined and also define different attack points in the
biometric system. The performance of the biometric system is evaluated as per ISO/IEC
19795 (2005) Standard. The main focus of this study is: Biometric template protection
schemes. fuzzy commitment schemes, fuzzy vault and homomorphic Bio-Crypto systems are
discussed in detail. All these schemes surely fulfilled the requirement of ISO/IEC 24745 i.e.
Recoverability, Irreversibility and Unlikability.
Keywords: Biometric, Attacks, Template protection schemes, Performance.
Introduction: The Biometric is the combination of two distinct words: bio and metric. The
term bio is related to human and the term metric is used to measure the characteristics of the
parts of a human body. With the advancement of technology, the biometrics are new began to
be used for security purposes. The characteristics of the human body are used for
identification, verification and authentication. It’s divided into two categories, based on
physiological and behavioural qualities. Physiological qualities are related to the shape of the
specific body parts and behavioural aspect related to the behaviour of the human beings, who
draw the pattern through actions. The identification system is based on three main
mechanisms, firstly the token based are tokens or Electronic keys, then the knowledge based
which are personal identifications and passwords, and lastly the biometric[1] as shows below
in fig 1

Identification Mechanism

Token Based Knowledge Based Biometric Based

Fig.1: Classification of identification mechanism
A combination of more than two mechanisms is referred to as multimodal and multifactor
authentication. The first two methods are not safe and reliable. The most reliable methods to
verify the identity of a person is biometric method, because every person has unique bio-
characteristics [2]. Biometric is related to Physiological characteristics like fingerprint, finger
vein, palm print, dorsal hand vein, retina, iris, ear and DNA[3], the use of tongue in
identification is under research. These characteristics of human varies from person to person,
even twins. And do not change during the entire lifetime. The biometrics are easily used with
the traditional systems for example: in financial organizations i.e. banking systems. The
fingerprints, iris, dorsal hand vein[4] etc. are used in place of a PIN, during the withdrawal of
money from ATM machines. These methods also apply in card less transaction for more
reliability and security. It may work separately or in combination with the PIN to securely
categorize user as genuine account holder in the bank. In India, ATM machines with
behavioural, and physiological identification are installed in utter Pradesh[5]. The structure
that works behind the Biometric schemes is to collect and store the data and to verify a
person’s identity. The biometric data and biometric identification technologies create the
biometric security systems. The verification process goes through three stages: image
capture, feature extraction and matching. During this process the biometric information
travels from one place to another. Thus, the security walls are built through various
cryptography algorithms[6]. Iris recognition is an automatic approach for identification. Iris
has a complex structure of vein in human eye, which differ from person to person.[7]
Background: The biometric system is the successor of Bertillon system, which was invented
by French criminologist, Alphonse Bertillon in 1879. It was a technique to describe
individuals based on the characteristics of their physical measurements. This included
standing height, sitting height, differences in the fingerprints. Alphonse Bertillon[7] was a
French police officer and a biometric researcher who applied the anthropological techniques
to identify the criminals. This is the first scientific system which is used by the police for
identification[8]. This system consists five different measurements: Head length, Head
breath, length of the middle finger, length of the left foot and length of the cupid. Although
the system was based on the scientific measurements, it was also known to have its flaws. It
may not have been able to accurately apply to children and women, as it was mostly designed
for men who had reached full physical maturity and had short hair.
Drawback of Bertillon system:
1. Women and children body different from time to time.
2. The devices used for measurement were necessarily of delicate construction and were
likely to get out of order.
3. Inaccuracy for identification.
Sir Galton provided the first practical fingerprint classification system after the Bertillon
system, which was later adapted by E. R. Henry is known as Dactyloscopy for practical
routine in police forces and many bureaucratic settings. The first fingerprint identification
system come into existence after 1970. This system is called a zeroth-generation
biometric system. This system was a stand-alone and has low performance. We can
divide the biometric technology based on the generation according to technology
advancement[9].
 The first-generation system has different types like face, fingerprints and iris recognition.
The applications developed based on the biometric technologies and are used in the
commercial and civil purposes. Various challenges faced by this generation are privacy,
security and performance related issues. In second generation the various new
technologies are developed based on the signal and image matching template and score
for identification and verification.[10] This generation also faced various challenges in
term of security the while transferring the biometric data over the communication
channel, in speed while identification and in authentication, efficiency of the system and
size of the application where biometric method are deployed all these challenges come
under the engineering perspectives and in social point of view the challenges are related
to protection policies of human body information while interacting to sensor, ethics
concern and health bias.
Now the biometric technology is growing well and used in many fields for security purposes.
The process of biometric technology goes from four stages, these are identification,
recognition, authentication and verification[11] for developing secure systems, these are
described in the fig. 2. An identification process system follows the 1:N model for
determination, the system compares the whole template stored in system with captured image
of biometric for true match, The verification possess the 1:1 model for comparison. To
validate the identity, the system compares the selected template in the identification process
with captured biometric data. The process is defined in fig. 3(a, b)

Identification: not confirmed Recognition process is used to

persons genuine or not just identify from the previous stored Closely related to
someone or something is identified knowledge in the system Identification

Authentication is the process of true Verification is the evidence which

Closely related to
match with the stored information confirm the accuracy of truth of
Security
in system identification

Fig. 2: Stages of biometric technology.

Process of Biometric technology: The process depends on two main parts: template
generation or Enrolment for template and template matching while authentication. During
template generation, the image captured by the device is considered as gallery image, then it
is send to feature extraction module for extraction and development of the reduced set
features is called feature vector (FV). Depending on this feature vector the template is
generated. The template is stored in the permanent storage for further use. In this phase the
identification is done. On another side: when the new query is assigned for authentication,
The image is captured and is sent to the feature extraction phase. The vector is developed on
live query basis. Finally, it sends to the classification phase for comparison. In this phase, the
verification is done. The classification of the biometric process is shown in the fig. 3(a, b).
While comparing, the similarity score is generated for each comparison based on these
scores, the system decides it is matched or not. The identification phase is more time
consuming and expensive during the term of computation, Because it needs 1: N matching.
Each live query finds its match in whole database’s templates. The phase of the verification is
less expensive than identification. The process of verification follows 1:1 matching of
selected templates while identification is verified based on the score matching to implement
the biometric process for security.[2] The feature extraction is the part of the pattern
recognition, which has faced more challenges during implementation. The accuracy of the
system directly depended on the

Template Template stored in

created permanent storage

Gallery Image Feature Vector

Capturing Device Feature Extraction Algorithm

Feature Vector
for live query
Classification
module
Live Query

Y/N

Template generation Template Comparison

Database of
Template 1:N
N Template

Identification Live Query Match

Selected Template 1:1

Verification Live Query Match

Fig. 3(a, b): Process of Biometric.

feature extraction algorithm. We can classify the feature extraction process in three broad
parts[12]:
• Local
• Global
• Region Based.
Some of the biometric attributes are classified as biometric system characteristics and are
defined in table 1: Applications are developed based on the uses and the target population is
divided into three categories[13]: Screening applications (to check the wanted person):
screening is the process of finding the wanted person who is living in the public space[7].
The list of the identities is developed as a watch list. The list includes the fewer identities.
This type of applications is developed to stop the unlawful activities. Implemented at airport
security[14], for terrorist identification and provides security at public places. The
applications are developed based on the single match i.e.: Verification system, which ensures
the authentication. These types of applications are used in commercial space such as: E-
transaction, system login, security of electronic data, ATM’s, mobile phones, medical
records. And in identification applications a large number of sample images are given to
system as input. When a live query comes to the system, the system finds the associated
pattern. These type of application include: national ID, voter ID, driving license and criminal
investigation[7][11].
Characteristics[7][2][11] Description
Universality Characteristics possessed by all human beings
Distinctiveness Each person has different characteristics in term of behaviour
and physiologicaly
Invariance The feature attributes of biometric images must invariant over
time.
Collectability The data should be easily collectable in every phase.
Performance Performance should be high on the term of accuracy.
Acceptability Population should be ready to feed the bio-information in
recognition system.
Circumvention Easily available to each user and has preserve the identity of
each user.
Simplicity The system should be simple enough so that every non-technical
person can use it
Cost Effective Cost should be low for designing and implementation.
Security The system should secure against fraudulent attacks
Privacy Information should be private for authenticate users of the
system
Security of the biometric system: The security of each system is major concern. When we
choose the biometric in another application to enhance security. Firstly, it should be clear, all
of the phases of technologies are secured in the term of data privacy and protection. The
biometric technology also faces various security attacks, The good feature extraction system
is responsible for higher security. On another side, the restricted system is responsible for
security attacks. The attacks use the fake biometric template generation and modification at
different levels, while identification, authentication and verification. The secured biometric
template protection scheme has followed four properties as given below:
 I. Diversity: To preserve the privacy of the template, while cross matching.
II. Revocability: in case of any damage, it should be possible to revoke the damaged
template and issue the new template based on its previous template features.
III. Security: it should be protected from any type of forgeries of templates.
IV. Performance: performance of matching template should be high
The biometric systems are much more sensitive because of security failures i.e.:
Vulnerabilities. Which are reasonable for a damage and offends to biometric template. The
different vulnerabilities exist, according to circumstances: Intrinsic non-fulfilment: when
protection failure occurs in the system, due to incorrect perception, because of false
interaction with the sensor device and the other reasons such as: diseases: for example, the
fingerprint sensor does not accept the quality image due to dry or wet skin type. The falsely
rejected and falsely accepted errors are occurring in this phase. When the chance of occurring
both errors is higher, then the zero-day assault attack takes place. We can stop this type of
attack by developing the more secure sensor., Administratively Advantaged: sometimes, the
body parts which are directly interacting with the sensor surface such as the fingerprints has
cut marks and the face has mark due to injury so, at the administration level privileges given
to system to register such type of users. It is very harmful, because any attacker can use these
privileges to register as genuine user. So, we need to keep the administrator privilege secret
and monitor each activity at the device interface, Admittance to biometric data: at the time of
the system development, the database developed by human being’s bio information the users
have not knowledge about this capturing. The data can be used for many unlawful activities,
non-secure foundation: the foundation includes the biometric components while designing as:
hardware, software, and communication channel for information transfer from one place to
another[15]. In whole foundation there are eight different points, where the attackers can take
an access of the system as shown in the fig.: 4
1. Fake biometric: feed the fake bio-information at the sensor surface. for example: fake
fingerprint, copy of signature etc.
2. Resubmitting stored bio-information: the previously stored bio-information
resubmitting by replacing the recording signals.
3. Overriding feature extraction process: this is done through trojan horse, attackers
generate the feature extraction sets, which is pre-selected by the intruders.
4. Altering the feature vector: the attackers place the new feature set in place of the
originals of the input signals. Two conditions occurre while transferring the input
signals for feature extraction: local and remote. When the template transfer over the
internet, the attackers can alter the certain packets.
5. Manipulate matchers: the attackers corrupt the functionality of the matcher module.
So that, it always produces the same preselected match score.
6. Modify template dataset: the database is either centralized or distributed over the local
or remote servers. Using this, intruders register themselves as genuine users. Smart
cards are mainly vulnerable to this type of attacks.
7. Attacks on channels: the channel used to transfer the information from the matcher
module to store module. The intruders can also modify the information over the
communication channel.
8. Modify the final decision: the intruders can attack the final decision. Even the whole
functionality is genuine or attack free.
In many ways the different bio-recognition system (Physiological and Behavioural)[2]. The
physiological recognition system is divided into three parts: Hand Region, to gain the access
of the fingerprint and the palm print recognition system, the attacker uses to make the fake
finger. For example: silica, gelatine, latex etc. The fingerprints collected from the various
sensor surface and other mediums where the victim has touched. In face region: the
recognition systems like face, ear, lips etc. The photos of the face are easily available to the
intruders from the social media. So that, attackers can easily hack the face recognition

5
3
8

Sensor Feature Vector Matcher Y/N

7
4
2
1

Template
6 database

Fig: 4. Attacker’s access points in biometric systems[15].

system. In ocular region: the iris and the retina mainly used for recognition system[2]. With
the advancement of the technology, high resolution systems are existing. So, it is very easy to
gain the access of the biometric recognition. But these types of attacks are much more
expensive due to high end optical system. And behavioural: the key strike pattern and touch
dynamics are difficult to copy from other user’s behaviour, but this is vulnerable to static
attack. The voice is travel in omni direction, so it is possible to record the voice of any
person. The attackers can also record the ECG signals[16] from Infrared sensors, but this is
easily detected and prevented. List of biometric security attacks, see table 2. List of
biometrics’ template attacks.
Name of Attack[10] Security breaches, Attack points
according to circumstances
Spoofing Adversary User interface
Zero-Effort-Attack Intrinsic failure User interface
Replay Attack [17]and Brute Non secure infrastructure Interface B/W module
Force
Hill Climbing[18] Non secure infrastructure Interface B/W module
DOS Biometric system failure Channel B/W module
Trojan Horse Biometric system failure Attacks at S/W module
Function Creep Non secure infrastructure Template’ database
Some Common Attacks of Non secure infrastructure Server
N/W[17]
The less secure and the vulnerable biometric system is prone to major security attacks various
security breaches is responsible for the same. The spoofing attack is also known as
presentation attack[19]. Because, the attackers change and replace the appearance of the
biometric samples. Zero-Effort-Attack, the occurrence of this attack is high when false accept
and false reject rate of the system is at highest, the low-quality sensors at recognition modules
are responsible for this fissure. No special efforts made by the adversary to cheat the system
(we can divide the adversary into three main categories: a) Administration attack: improper
system administration. b) Non secure infrastructure: h/w, s/w, and communication channels
between modules. c) Biometric overtness: sensors are not able to detect the liveness). Replay
Attack: During transmission, the attacker sends the old data of the authorized users. Which is
stolen by the attackers to impersonate a victim. A hill-climbing and brute force [20]attack
might be achieved by an application, That sends random templates to the system, which
disturb the system functionality iteratively. The application recites the output match score and
continues with the perturbed template. Only when the similarity score increases until the
conclusion threshold is topped. DOS (denial of service) it is occurring when system denies to
authorize users for service delivery because the adversary change the template which is
stored in the system or totally corrupt the system, not any user can take the advantage. Trojan
horse: sent the executable program by the adversary to disturb the software functionality, So
that, the adversary can get the desired output. Function creep: the attacker steals biometric
data and use for other inadvertent purpose. The adversary can harm the system by some other
simple network attacks like: SQL injection, power lifting, hijacking. The two main problems
faced by the biometric system are compromising template will not be cancelled and reissued
like password and token. The one solution to stop the fake biometric, the concept is liveness
detection.[12]
Performance measurements: the performance of any system directly depends on its Error
rate, which are occurring at various levels of the system development. The error rate
iscalculated for each activity during the whole process. The accuracy of the system, mainly
depends on the performance: capturing a quality image, short time interval between the
enrolment and verification phase, robustness of recognition system, and the environmental
factors: temperature, humidity, illumination conditions of the sensor surface[7]. To evaluate
the performance of the biometric system, all criteria are defined in the standard way in the
series of ISO/IEC 19795 (2005). The accuracy is defined in the terms of error rate as:
I. Sample acquisition error: This type of error arises while capturing the image. The
terms defined in as: failure to enrol (FTE): system rejects the sample. And failure to
capture (FTC): inability of the sensor to capture the image.
II. Performance error: it is used to calculate the accuracy of the system in the real world.
This is also divided in the two sub errors:
a) False match rate (FMR) or false accept rate (FAR): this is called type 1 error. This
type of error is often in the system, because most of the systems are highly
accurate. The error is defined as the percentage of invalid matches.
𝐹𝐴𝑅 = 𝐹𝑀𝑅 ∗ (1 − 𝐹𝑇𝐴)
b) False non match rate (FNMR) or false reject rate (FRR): the error is defined as
reject the authorized users. It is called type II error.
𝐹𝑅𝑅 = 𝐹𝑇𝐴 + 𝐹𝑁𝑀𝑅 ∗ (1 − 𝐹𝑇𝐴)
III. Equal error rate or accuracy: in this condition both false reject rate and false accept
rate are equal. This is done with ROC curve plot graphical representation. Plot the
false accept rate against the false reject rate. This trade off describes the threshold
value. We can achieve the higher accuracy by decreasing all mentioned errors:
𝐹𝑁𝑀𝑅 = 𝐹𝑀𝑅
In case of the similarity score FRR and FAR are calculated in pairs. The result is
presented in many ways: % (percentage) 1/100, in decimal form and in the power
of the 102 while comparison. The more accurate system has lower FRR at the
same point of the FAR. In case of the non-similarity score, the system takes
decisions based on match and non-match single pair of FAR / FRR pairs
IV. Identification rate (IR): return the correct identify user in proposition to the enrolled
users.
a) False negative identification error rate (FNIR): In this, the system is failing to
returne the genuine users who have enrolled. This is defined in the term of a
proposition. The 1:N system of verification is defined as:
𝐹𝑁𝐼𝑅 = 𝐹𝑇𝐴 + (1 − 𝐹𝑇𝐴) ∗ 𝐹𝑁𝑀𝑅
b) False positive identification error rate (FPIR): In this, the system is not able to
enrol the attempts by the identifiers. One attempt made against N size of the
database as:
𝐹𝑃𝐼𝑅 = (1 − 𝐹𝑇𝐴)(1 − (1 − 𝐹𝑀𝑅)𝑁 )
V. Genuine accept rate (GAR): This is measured in the term of percentage of the
registered users, identified by authentication infrastructure is calculated as:
𝑛𝑜. 𝑜𝑓 𝑔𝑒𝑛𝑢𝑖𝑛𝑒 𝑢𝑠𝑒𝑟𝑠 𝑎𝑐𝑐𝑒𝑝𝑡𝑒𝑑
𝐺𝐴𝑅 =
𝑛𝑜. 𝑜𝑓 𝑔𝑒𝑛𝑢𝑖𝑛𝑒 𝑡𝑟𝑎𝑖𝑙𝑠
The performance is represented using the graphical representation using. ROC curve
(Receiver operating characteristics) used to plot the true positive rate (1-FRR). At the top
of this plot, we can consider the best biometric system against false positive rate, DET
(Detection error trade off) is also used to plot FRR rate against false accept rate. Both
plotting methods are used for binary classification. The ROC curve is used to measure the
1:1 verification system, and CMC (cumulative match characteristics curve) is used to
define the 1:N system of identification. In this correctly identified users plot on the Y axis
and the rank value of the X axis, it is defined in the term of probability[2][10][21]

Security Schemes for biometric’ template protection

The security of the biometric template is different from the password security while
authentication. The reason behind this is that the, users easily recovers the password and set a
new password for every authentication attempt. The techniques used for password recovery
totally follow a different scenario than biometric template protection. The ISO/IEC 24745
Standard[21] for biometric complete protection provides the recoverability, irreversibility and
unlink-ability[22]. The standard enhances the security and privacy of the templates.
1 Recoverability: In this, users are able to generate the multiple reference template from the
same biometric information. This property reduces the risk of a compromised template. It
ensures the newly created template should not cross match with all other previously created
templates when it is Revoked for authentication.
2 Irreversibility: The newly referred template is protected in such a way that any Intruder is
not able to gain access.
3. Unlink-ability: Different templates generated from the same biometric information traits is
Unlink-ability.
Fuzzy commitment Schemes[15]: The fuzzy logic is a way to define the approximation reasoning in
the real world. It is based on the degree of probability of the truth value instead of true or false
diseases. It is a part of artificial intelligence. The commitment scheme is the combination of the two
techniques in information security. When We combine the digital key information to a biometric
pattern, we can call it a fuzzy commitment scheme[23]. The two techniques are error correcting codes
and Cryptography. In this, the users can convert the value in more than one way, so that the intruders
cannot reverse it.
Diagram
[23] Ari Juels et al. (1999) proposed a new technique i.e. Fuzzy commitment. Which is based
on error correcting codes and Cryptography. It works on noisy channels. It overcomes the
major problem in biometric authentication. This technique can be achieved with three
different roles such as offline authentication, challenge response authentication and
encryption decryption. The experiment is done on fingerprint database. [24] Yevgaiy dods et
al. (2004) This technique is proposed not only for biometric key generations for security but
for any data which has been need of keying security. The fuzzy extractor is used to collect the
randomness of selected biometric information, which was used as key purpose to protect the
biometric templates. The key remains unchanged until the change in original pattern.
[25]Christan Pathgeb et al. (2009) The research proposed the iris-based commitment
schemes. In this, the texture of the Iris is treated as a transient signal after that, is processed
by wavelet transformation. The Hadamard codes were applied for error correction. The
experiment is done on CASIA- IRIS V3 interval database. FAR and FRR are 0.08% and
6.57% respectively. [26] C. Rathegeb (2010) Propose a new method for iris based fuzzy
commitment schemes. Which is based on the inter-class error code analysis. The iris code
rearranges in such an away so that, the system achieves the higher accuracy rate. Performance
measured in the terms of false accept rate and false reject rate (FAR and FRR). The main
focus of this method to enhance the efficiency of the fuzzy commitment schemes. Error
correction decoding done over CASIA V3 iris database. [27]Marta Gomez et al.(2016)
Proposed a method based on Bloom filters for unlink-ability and irreversibility of the
biometric templates. The experiment was done on a bio-secure multimodal database which is
used in previous research work. The performance is measured in terms of FNMR, equal error
rate of 6.25% is obtained for different configurations. [28]Andrew beng jin toeh et al. (2017)
the researchers proposed randomised dynamic quantization transformation for binary data of
bio information. In this, bit strings are highly distinctive among all the users while
transferring. The experiment was done on fingerprint database FVC 2002. This method
satisfies the uniqueness and randomness properties of the biometric standard. [29]Thi ai thao
Nguyen et al.(2019) Proposed hybrid technique using random orthogonal projection for
fuzzy commitment schemes. This is developed to secure the invertible cryptographic keys
for cancellable templates. The accuracy of the system is measured in the terms of error rate
FAR FRR and EER as 9 % all values .Accuracy of the system was achieved 91%.
[30]Shivangi Shukla et al.(2021 march) In this, the new method is developed based on fuzzy
commitment schemes by enhancing the minutiae point in fingerprint templates. FVC 2000-
DB2, FVC 2002- DB1, FVC 2002-DB2, FVC 2004-DB1 were used for experimental
purposes. The Bose Chaudhuri- Hocquenghem
Table 3. Summarization of Fuzzy Commitment Scheme.
Author Year Technique Efficiency
[23]Ari Juels et.al. 1999 Developed new technique: Fuzzy
Commitment Scheme
[24]Yevgeniy Dodis 2004 Modified JS secure Sketch & BCH
et.al. (Binary Coded Hexadecimal based
secure Sketch)
[25]Christain 2009 Iris based fuzzy commitment scheme Block level (80-16)/2=32
Rathegeb et.al. FAR=0% & FRR=4.64%
Block Level (250-16)/2=117
FAR=0.08% & FRR= 6.57%
[26]C. Rathgeb 2010 Rearrangement of bits to achieve the FRR=4.92%
et.al. efficiency of error correcting decoding
(Hadamard Decoding)
[27]Marta Gomez- 2016 Bloom Filter EER=1.3%
Barrero et.al. With random Shuffling EER=40%
[28]Andrew Beng 2017 RDQT (Randomized Dynamic FRR According to parameter setting (15,7)=0.9%, (15,9)=1.7%,
Jin Teoh et.al. Quantization Transformation) (15,11)=2.3%
[29]Thi Ai Thao 2019 Random Orthogonal Protection and FRR, FAR, EER =9%
Nguyen et.al. fuzzy commitment Scheme
[30]Shivangi 2021 BCH Code for error correcting SHA- FVC2000-DB2 FRR=2.5% & FAR=0.29%
Shukla et.al. 256 for security FVC2002-DB1 FRR=3.5% & FAR=0.89%
FVC2002-DB2 FRR=4.5% & FAR=0.98%
FVC2004-DB1 FRR=5.5% & FAR=0.84%
[31]Alawi A. Al- 2021 PQFC (Post Quantum Fuzzy Accuracy= 99.1 %
Saggaf et.al. Commitment) Scheme FAR=0% & FRR=2.9%
[32]Tim Van 2021 Security Analysis based on the inertial Temp- HMM HMM ANN ANN Fusion Fusion
hamme et.al. measurement unit (IMU) gait late FAR% FRR% FAR% FRR% FAR% FRR%
authentication, HMM, ANN, AND 63 0.05 6.21 0.22 4.35 0.06 5.38
Combination. 127 0.03 6.63 0.11 4.55 0.01 8.70
255 0.02 8.07 0.08 4.14 0.02 6.83
511 0.17 2.28 0.07 4.55 0.01 7.25
(BCH) is used for error correction and secure the codeword by use of SHA-256 Hash
Function. The efficiency was measured as FRR and FAR. [22]Alawi a. saggaf et al.(2021
july)Drive the new post-quantum fuzzy commitment schemes. This method is used to
overcome the short factor problem of lattice (SVP). The Iris database of 108 samples was
used in this method for experiment and the accuracy of the system is 99.1% achieved. FAR of
97.4 % and FRR is 2.9% achieved, without post-quantum Fuzzy commitment schemes..
[32]Tim when Hema met al.(2021 nov 12) In this study, presents the Security Analysis which
is based on the inertial measurement unit(IMU) gait authentication. Two models were
developed to evaluate the effectiveness of the predicted and unpredicted system. The analysis
was done over OU- ISIR data set which is labelled as level walk.
fuzzy vault scheme: The cryptographic technique, in which traditional cryptographic methods
are combined with biometric authentication is called a fuzzy vault scheme[33]. This scheme
is mainly used to provide security to keys used in biometric template protection as same as
advanced encryption standard (AES). The procedure of the fuzzy vault locked the template of
the biometric using a key.
1. Salting, 2.Non invertible transform
In salting, the transform template is invertible which directly depends on the key's secrecy and the
non-invertible transform. In this, it is very hard to revert the transformed template of biometric. We
know that, the helper data is the combination of the biometric template and keys, which is used for
building the key binding biometric cryptosystem. The matching function is performed to recover the
key indirectly from the helper data by applying query bio-metric features.
Ari juels et al.[34] (2005) Developed a new technique which is based on the traditional and
new cryptosystem called the Fuzzy vault. The helper data is created by binding the key with a
template of biometric. In the Real world, the user of the fuzzy Vault is in error-prone
channels while transferring information. The experiment was done on a fingerprint database.
The locking and unlocking algorithm of the fuzzy vault is used in the whole process for
protection. This is based on Reed Solomon codes. The fuzzy vault scheme is the successor of
the fuzzy commitment scheme. [35]Umut uludag et al. (2005) Author presents the method
which is based on the fuzzy vault for fingerprint minutiae features. The IBM-GTBD
fingerprint Database used for practical purposes. The proposed architecture was able to
secure the same as AES 128-bit key algorithm. The drawback of this technique, the time
complexity is very high, because multiple points are evaluated during coding. [36]Li Qiang et
al. (2006 a) The author improved the previous technique, the Fuzzy vault. The unlock
algorithm in this technique has suffering from time complexity issue which was based on the
Reed Solomon code. Now, The Hash Function is used to store the biometric information in a
vault. [37]Jason jeffers(2006) In this study, the author investigates the three different
minutiae representations for fingerprint biometric. The data set was created by the researcher
of Thumb and fingerprint of 37 human beings. Initially voronoi and triangular based
structures was used for translation and rotation for bio cryptographic system. The voronoi
representation has a higher performance than others representations. [38]Youn joo lee(2007)
the researchers developed the new method to handle the problem of variation in feature
extraction over the Iris database BERT Version1. The feature extraction algorithm is based
on independent component analysis (ICA). The 128-key is used in cryptosystems. The
accuracy of the system achieved in the term of error rate as: FAR 0 % and FRR 0.77%,GAR
99.225%. [39]Karthik nandakumar(2007- 4-dec) In this article, developed method, which is
used to align the query finger print with its template using fuzzy vault. The highly inflected
points in the fingerprint orientation were extracted and used for template alignment i.e. helper
data. The two different databases used FVC 2002-DB2 and MS-DBI of fingerprint. [40]E.
srinivasa reddy et al.(2008)It represents the new method to overcome the problem of
nonuniform nature and cross matching of biometric data. The developed method was based
on hardness i.e. add another layer of security, Iris structure-based password is used to secure
the fuzzy vault and key for two different databases CASIA and MMU of iris. [41]Karthik
Nandakumar (2008) In this, the developed method is used to protect the multiple templates as
a single entity of a user. It provides higher efficiency than a uni-biometric model. The multi
biometric Vault method applied on fingerprint and Iris has achieved a rate FAR 0.01% and
GAR 98.2 %. [42]Eryun liu (2009) Proposed more feature based key generation for
fingerprint biometric to achieve the higher accuracy than the previous Developed model.
Under the fuzzy Extractor framework, the researcher proposed the fingerprint-based
generation of key for security. In this, two types of features and 3 types of sketch were used
for feature differences. The FVC 2002-DB1 and DB2 were used for the experiment.
[43]liafang wu (2010) In this, proposed system overcomes the problem of the diversity and
recoverability of the Fuzzy wallet. The experiment is done with outline face based
authentication. In this, a transfer form template is used which enables the system for diversity
and revoke ability. Principal component analysis (PCA) is used as a feature extractor. the
CRC 144-bit of user key for face ORL database. [44]Mohamad khalil hani(23 feb 2012)
Developed algorithm for biometric encryption to reduce the computational time i.e. fast Chaff
generation algorithm. It enables the system to implement the encryption process of biometric
on stand-alone device i.e. system on chip. [45]Julier bringer (2012) Proposed the extension to
store the multiple user biometric information which is able share the same vault. In this,
locked the fuzzy vault by different sets such as : Ai= 1……j .It is based on the flooded reed-
solomon code. It is the method which support parallel users of Single fuzzy vault. [46]Ki
young moon(2012)Propose three different solutions for fuzzy fingerprint vault. Automatic
alignment, resistant to correlation attack and OTT (one time template). The FVC 2002
database is used for implementation; the geometric hash table method is used for automatic
alignment. [47]chriatan rathgeb (2016) The proposed technique based on two-dimensional
binary feature extraction vectors, for Iris code from the Iris images. The iris code is divided
vertically. The experiment was done on the CASIA version 3 intervals and IITD version 1
Iris database. The efficiency of This solution mapped in the form of GMR 95% and
97%.[48]girish revadiger (2017) secure key generation for groups using acceleration and
fuzzy vault for smart wearable devices. The routine activities were considered as Z to
generate the key.[49] V.sujitha (17 feb 2018)Developed the new method using fuzzy vault,
the feature extraction of the CASIA Palmprint database using bottom hat Filtering. To create
the projection point in biometric information, the polynomial is used. When the matching
score is greater than the predefined value of Threshold the Vault successfully decodes. during
experiment, security is analyse for brute force attack.[50] Katarzyna kaptyra ( 21 feb 2018)
Proposed multi secret fuzzy vault. In this, all piece of information stored in a single wallet
and each piece has its own different key. During the recovery process, only used key
information is recovered. This is the improved version of base Fuzzy vault. [51]Sweedle
machado (2018) proposed a Procedure to protect the ATM (automatic teller machine)’ pin
and password using bio-information i.e. fingerprint. The ATM pin and password encoded and
stored in the fuzzy vault either on mobile, smart card etc. to revoke the PIN and password
using Fingerprints’ biometric information. the experiment done on the dummy fingerprint
database of 80 images.[52].lin you et al. (2018 14 march) Proposed advanced version of
Fuzzy Vault i.e. a single template store in a vault. The proposed scheme is based on the
multiple Bio template and non-stored feature points in fuzzy Vault due to security reasons. In
this, fingerprint and finger vein feature fusion are used to overcome the limitation of
traditional vault. Database created by the laboratory independently. The grid projecting
method is used for feature point generation. And the Efficiency of schemes Me
Table 4. Summarization of Fuzzy Vault Scheme.
Author Year Technique Efficiency
[34] Ari Juels et.al. 2005 Developed new technique: Fuzzy Vault Scheme (Reed-
Solomon Codeword)
[35]Umut Uludag 2005 Minutiae features of fingerprint dataset FAR=0%
et.al.
[36] LI Qiong et.al. 2006 New UNLOCK Algorithm for fuzzy vault
[37] Jason Jeffers 2006 Initially Voronoi and triangular based structures FRR=4.92%
et.al.
[38] Youn Joo Lee 2007 Developed new algorithm based on independent FAR=0%, FRR=0.77%, GAR=99.225%
et.al. component analysis
[39]Karthik 2007 Iterative closest point (ICP) based new method for finger print. [39]
Nandakumar et.al.
[40]E. Srinivasa 2008 Add another layer of security using password Degree of GAR% FAR%
Reddy et.al. polynomial
4 84 0
5 87 0
6 89.7 0
7 90.4 0
8 90.2 0
[41]Karthik 2008 Multi-biometric vault method FAR=0.01%, GAR=98.2%
Nandakumar et.al.
[42]Eryun Liu 2009 Feature Based Key Generation [42]
et.al.
[43]Liafang Wu 2010 Fuzzy vault scheme for online Authentication [43]
et.al.
[44] Mohamad 2012 Fast Chaff Generation Algorithm
Khalil Hani et.al.

[45]Julien Bringer 2012 Flooded Reed Solomon Code

et.al.
Author Year Technique Efficiency
[46]Ki Young 2012 3-Different solution as: Automatic Alignment, Resistant to On 7 % Degree of polynomial FAR=0% and
Moon et.al. Corelation attack, OTT (one time template) GAR=92.1%
[47]Christian 2016 Two- Dimensional Binary Feature Extraction Vector GMR=95% and 97% at single and multi-
Rathgeb et.al. Instance Respectively
[48]Girish 2017 Secure Key Generation for Smart Wearable Devices using
Revadigar et.al. Fuzzy Vault
[49]V. Suitha et.al. 2018 Palmprint based fuzzy vault scheme FRR=0% and FAR=0.02%
[50]Katarzyna 2018 Multi- Secret Fuzzy Vault developed to improve the base
Koptyra et.al. fuzzy vault scheme
[51]Sweedle 2018 Fuzzy Vault for ATM credential
Machado et.al.
[52]Lin You et.al. 2018 Advance version of fuzzy vault i.e., store only single GAR=95% and FAR=0%
template in vault. Grid Projecting method used for feature
Extraction
[53]V. Sujitha et.al. 2019 Bottom Hat Filtering GAR=95%
[54]Wendy Ponce- 2019 Analyse 2 different methods of Fuzzy Vault [54]
Hernandez et.al.
[33]Raza 2020 Polynomial based Fuzzy Vault GAR=92%, FRR=90%, FAR=85%
Mahmood et.al.
[55] Wendy Ponce- 2020 FMR of 6.91% and FNMR of 7.85% using
Hernandez et.al. MCYT database, FMR of 6.21% and FNMR
of 4.86% for proprietary database, and FMR
of 6.16% and FNMR of 13.6% for Bio-
Secure database
[56]Xingbo Dong 2021 IoM (Index of Maximum Hashing) & DNN (Deep Neural
et.al. Network)
[57]Vivek Singh 2021 PCA based alignment for template and prob image [57]
Baghel et.al.
 (a)

(b)

Fig: (a,b) Vault Encoding and Decoding of Fuzzy Vault Scheme[52].

asured in the term of GAR and FAR are 95% and 0.4 % respectively. [53]V. sujitha (23 dec
2019) Proposed the new technique in which multiple biometric information such as Palm
print and fingerprint template are used. to improve the security of template. Image
enhancement, Image segmentation and Bottom Hat Filtering are used. This system has a high
GAR rate to handle brute force attacks.[54] Wndy ponce hernandez (2019)In this study, two
different methods of Fuzzy vault are analysed, first is fuzzy vault for cyclic redundancy
check(CRC). The computational cost of this method is high due to no error detecting process
during encoding. In another one, per query verification is performed, this variant is analysed
when interpolation operation and error correcting operation are performed separately.[33]
Reza mehmood (6 jan 2020) Proposed method based on polynomials to hide the key and
transform it by using an integral operator. the key is not available for long period of time, if
the attacker owns the polynomial. This is an advantage of this proposed method. This scheme
overcomes the problem of key stolon attack and provides the higher efficiency in terms of
GAR, FRR, FAR ,92%, 90%, 85% of GAR for 3,4,5 polynomial representation. The
experiment done on FVC 2002 -DB1, DB2,DB3,DB4 SET OF B database.[55] wndy ponce
(17 jan 2020)Developed the system based on fixed length templates for fuzzy vault. which is
applied on dynamic signature verification. To evaluate the performance of this system, three
different databases are used as: bio-secure, MCTT and propairitry collection for
signatures.[56] Xingbo dong(july 2021) Proposed the face cryptosystem for identification for
template protection which follows 1 to N model and for Secret protection 1 to 1 model is
used. Which is implemented by index of maximum hashing(IoM). The IJB-C, VGG2 and
LFW databases are used. To extract the feature of biometric information, the deep neural
network and fusion module for biometric are applied.[57] Vivek singh bahel(23 july 2021)
the proposed technique filter the point which is the combination of both genuine And chaft
point of the fuzzy vault. The PCA (principal component analysis) used for alignment of
template with gallery image. three different databases are used for the experiment FVC 2002-
DB1, FVC 2002-DB2 FVC 2004-DB1.
Homo-morphic encryption: Homomorphic encryption is derived from traditional
cryptography. The origin of the homomorphic Encryption is RSA. This is the first public key
encryption scheme; the random bits are padded into message losing the property of the
homomorphic encryption. Avoiding padding in the message results of multiple versions of
the Homo morphic encryption. Now a days this technique is combined with biometric
template to achieve the higher security[58].
[59]Wilson abel (2014) In the proposed method, the full homomorphic encryption version
was used for privacy preserving of the iris database i.e. BATH. the computation is done over
encrypted data which is preserving the privacy of biometric data. the technique is
implemented with Java free library as Java lattice-based cryptography. [60]Marta
Gomez(2016)The proposed technique meets the requirement of ISO/IEC 24745 standard for
protection of biometric that is irreversibility, renewability and unlinkability. This technique
used for fixed length template. In this, only encrypted data is handled. The cost of
computation is very low. This is implemented with an online signature. [61]marta
gomez(2017) The method based on multi biometric information of online signatures and
fingerprints. The database of bio secure ID Multimodal is used. This model is based on
probabilistic homomorphic encryption. Which is satisfied the requirement of ISO/IEC 24745
standard for biometric. The multiple models are analysed and describe three fusion levels as
described in the ITEC TR 24722 for multimodal. The efficiency measure in the term of EER
is 0.12%. [62]Andrew Nautseh (9 march 2018) Developed a method for bridging the gap
between complete protection for speaker recognition. The method is implemented using
paillier Cryptosystem, cosine comparators and Euclidean distance. Two architecture models
are built, first is to use for privacy of biometric data and second is to encrypt the comparator
of parameters. [63]P. Drozdonski (2019) Proposed the technique for special identification by
using two different homomorphic encryptions for biometric templates. Which are
implemented using the frontal subset of the FERET database. The two different variations of
the homomorphic are Brakerski/fan-vercauteren and cheon-kim-kim. This proposed system
fullfils the requirement of ISO/IEC 24745 Standard for bio- metric. [64]Mahesh kumar (23
nov 2021) Proposed the secure and verifiable machine learning based architecture system,
which is encrypted using full homomorphic encryption Method. This system is developed for
cloud servers. The CASIA- V3 interval and IITD are used for experiments. using private
nearest neighbour algorithms
Table 5.Summarization of Homomorphic Biometric Template Scheme.
Author Year Technique Efficiency
[59]Wilson Abel 2004 Full Homomorphic Encryption for iris
et.al.
[60]Marta Gomez 2016 Fixed Length Template protection based
Barrero et.al. on Homomorphic
[61]Marta Gomez 2017 Probabilistic based Homomorphic EER=0.12%
Barrero et.al. Encryption Multimodal
[62]Andrew 2018 paillier Cryptosystem, cosine
Nautsch et.al. comparators and Euclidean distance
[63]P. Drozdowski 2019 Brakerski/fan-vercauteren and cheon-
et. al. kim-kim.
[64]Mahesh 2020 Machine Learning based Homomorphic
Kumar Encryption (secure and verifiable [64]
Marampudi et.al. classification-based iris authentication
system)
[65]Ebenezer Okoh 2021 DNN and Homomorphic
et.al.
[66]Hiroto Tamiya 2021 Post Quantum and Packed It is 14.0 times faster than that of the HE-based face template
et.al. Homomorphic protection system presented in BIOSIG2020
[67]Mauro Barni 2021 Homomorphic Encryption for privacy
et.al.
[68]Amina Bassit 2021 Analysis of Homomorphic Encryption
et.al. and Bloom Filter
[69]Xiaopeng Yang 2021 FITing Tree, iDistance and symmetric
et.al. homomorphic encryption
and private multi-class perceptions are used for testing and training of the encrypted template.
[65]Edenezer okoh (2021) Proposed this system which is further divided into two modules as
first one is Deep neural network and second one is homomorphic encryption. The system
implement using mobile face images and periocular characteristics. [66]Hiroto taniya (2021)
The proposed system is the improved version of BIOSIG 2020 of face template protection.
The result of this development is 14 times faster than the previous developed model. The
squared euclidean distance (SED) is used to compute the encryption by a single
homomorphic encryption multiplication. Arcface database of face is used for experiment.
[67]Mauro barni(2021) this system is mainly developed for distributed biometric systems to
preserve the privacy of the individual using homomorphic encryption finger-code template.
After that the encrypted data should be shared with the server. The cross match data set of
fingerprints is used for experiment which is publicly available. [68] Amina bassit (2021) In
this article, analysed the two different schemes such as Bloom filter versus homomorphic
encryption are checked it out. Which one is better over the other? The author's analysis is that
schemes have strengths and weaknesses. but both schemes preserve the accuracy and fulfill
the requirements of the ISO/IEC 24745 biometric standard. [69] xiaopng yang (28 sep 2021)
Proposed a privacy-preserving information of biometric identification which is based on the
FITing tree, iDistance and symmetric homomorphic encryption for two cloud servers. The
computational cost of the cloud server for biometric searching is acceptable. The proposed
system ensures the privacy of the server provider and user (client) both

References:
[1] S. Sankar, B. Pankaj, O. Marques, B. Majhi, and S. Bakshi, “Iris recognition with
tunable filter bank based feature,” pp. 7637–7674, 2018.
[2] K. Shaheed et al., A Systematic Review on Physiological ‑ Based Biometric
Recognition Systems : Current and Future Trends, vol. 28, no. 7. Springer
Netherlands, 2021.
[3] S. Z. Li, “Encyclopedia of Biometrics.”
[4] W. Jia et al., “A survey on dorsal hand vein biometrics,” Pattern Recognit., vol. 120,
p. 108122, 2021.
[5] A. T. Siddiqui, “Biometrics to Control ATM Scams : A Study,” pp. 1598–1602, 2014.
[6] R. Karinattu and D. Job, “Design and Implementation of Secure ATM System Using
Biometric and Hashing Technique,” vol. 8080, pp. 0–2, 2021.
[7] J. A. Unar, W. Chaw, and A. Abbasi, “A review of biometric technology along with
trends and prospects,” Pattern Recognit., vol. 47, no. 8, pp. 2673–2688, 2014.
[8] “Assistant Physician Kankakee, LI.,” 1896.
[9] R. B. Fosdick, “Passing of the Bertillon System of Identification,” vol. 6, no. 3, 1916.
[10] A. K. Jain and A. Kumar, Biometric Recognition : An Overview. 2012.
[11] A. K. Jain, A. Ross, and S. Prabhakar, “An Introduction to Biometric Recognition,”
vol. 14, no. 1, pp. 4–20, 2004.
[12] K. Joshi and M. I. Patel, “Recent advances in local feature detector and descriptor : a
literature survey,” Int. J. Multimed. Inf. Retr., vol. 9, no. 4, pp. 231–247, 2020.
[13] A. K. Jain, S. Pankanti, S. Prabhakar, L. Hong, and A. Ross, “Biometrics : A Grand
Challenge.”
[14] K. O. Gordyushova, L. I. Rogavichene, E. V Budrina, and A. S. Lebedeva, “Biometric
Technologies : Application Possibilities At Airport for Passenger Processing,” vol.
137, pp. 49–52, 2020.
[15] A. Sarkar and B. K. Singh, “A review on performance , security and various biometric
template protection schemes for biometric authentication systems,” pp. 27721–27776,
2020.
[16] S. Bianco and P. Napoletano, “Biometric Recognition Using Multimodal Physiological
Signals,” IEEE Access, vol. 7, pp. 83581–83588, 2020.
[17] Z. Rui and Z. Yan, “A Survey on Biometric Authentication : Toward Secure and
Privacy-Preserving Identification,” IEEE Access, vol. 7, pp. 5994–6009, 2019.
[18] I. J. Jacob, P. Betty, P. E. Darney, S. Raja, and Y. H. Robinson, “Biometric template
security using DNA codec based transformation,” vol. 5, pp. 7547–7566, 2021.
[19] R. N. Rodrigues, L. Luan, and V. Govindaraju, “Robustness of multimodal biometric
fusion methods against spoof attacks,” J. Vis. Lang. Comput., vol. 20, no. 3, pp. 169–
179, 2009.
[20] J. A. Siguenza, “Hill-Climbing and Brute-Force Attacks on Biometric Systems: A
Case Study in Match-on-Card Fingerprint Verification,” pp. 151–159, 2006.
[21] T. Fundamentals, “UNDERSTANDING BIOMETRIC PERFORMANCE
EVALUATION.”
[22] R. Arjona and I. Baturone, “A Post-Quantum Biometric Template Protection Scheme
Based on Learning Parity With Noise ( LPN ) Commitments,” pp. 182355–182365,
2020.
[23] A. Juels, C. Drive, M. Wattenberg, and W. Street, “A Fuzzy Commitment Scheme,”
pp. 28–36.
[24] Y. Dodis, L. Reyzin, and A. Smith, “Fuzzy Extractors : How to Generate Strong Keys
from Biometrics and Other Noisy Data,” pp. 523–540, 2004.
[25] C. Rathgeb and A. Uhl, “Systematic Construction of Iris-Based Fuzzy Commitment
Schemes,” pp. 940–949, 2009.
[26] C. Rathgeb and A. Uhl, “ADAPTIVE FUZZY COMMITMENT SCHEME BASED
ON IRIS-CODE ERROR ANALYSIS C . Rathgeb and A . Uhl Department of
Computer Sciences University of Salzburg , A-5020 Salzburg , Austria,” pp. 41–44,
2010.
[27] M. Gomez-barrero, C. Rathgeb, J. Galbally, C. Busch, and J. Fierrez, “Unlinkable and
irreversible biometric template protection based on bloom filters,” Inf. Sci. (Ny)., vol.
370–371, pp. 18–32, 2016.
[28] A. Jin and J. Kim, “Secure biometric template protection in fuzzy commitment
scheme,” vol. 4, no. 23, pp. 724–730, 2007.
[29] T. Ai, T. Nguyen, T. K. Dang, and D. T. Nguyen, A New Biometric Template
Protection Using Random Orthonormal Projection and Fuzzy Commitment. Springer
 International Publishing, 2019.
[30] S. Shukla and S. J. Patel, “Securing fingerprint templates by enhanced minutiae ‐
based encoding scheme in Fuzzy Commitment,” no. July 2020, pp. 256–266, 2021.
[31] A. A. Al-saggaf, “A Post-Quantum Fuzzy Commitment Scheme for Biometric
Template Protection : An Experimental Study,” IEEE Access, vol. 9, pp. 110952–
110961, 2021.
[32] T. Van, E. A. Rúa, D. Preuveneers, and W. Joosen, “On the Security of Biometrics and
Fuzzy Commitment Cryptosystems : A Study on Gait Authentication,” vol. 16, pp.
5211–5224, 2021.
[33] R. Mehmood and A. Selwal, “Polynomial Based Fuzzy Vault Technique for Template
Security in Fingerprint Biometrics,” vol. 17, no. 6, pp. 926–934, 2020.
[34] A. R. I. Juels, “A Fuzzy Vault Scheme,” pp. 237–257, 2006.
[35] U. Uludag, S. Pankanti, and A. K. Jain, “A Fuzzy Vault for Fingerprints,” pp. 310–
319, 2005.
[36] L. I. Qiong, L. I. U. Zhaoqing, and N. I. U. Xiamu, “Analysis and Problems on Fuzzy
Vault Scheme,” pp. 2–5, 2006.
[37] J. Jeffers and A. Arakala, “MINUTIAE-BASED STRUCTURES FOR A FUZZY
VAULT,” 2006.
[38] Y. J. Lee, K. Bae, S. J. Lee, K. R. Park, and J. Kim, “Biometric Key Binding : Fuzzy
Vault Based on Iris Images,” pp. 800–808.
[39] K. Nandakumar, S. Member, and A. K. Jain, “Fingerprint-Based Fuzzy Vault :
Implementation and Performance,” vol. 2, no. 4, pp. 744–757, 2007.
[40] E. S. Reddy and I. R. Babu, “Performance of Iris Based Hard Fuzzy Vault,” pp. 248–
253, 2008.
[41] K. Nandakumar and A. K. Jain, “Multibiometric Template Security Using Fuzzy
Vault,” vol. 00, pp. 7–12.
[42] E. Liu, J. Liang, L. Pang, M. Xie, and J. Tian, “Journal of Network and Computer
Applications Minutiae and modified Biocode fusion for fingerprint-based key
generation,” J. Netw. Comput. Appl., vol. 33, no. 3, pp. 221–235, 2010.
[43] L. Wu, “A Face Based Fuzzy Vault Scheme for Secure Online Authentication,” 2010.
[44] M. Khalil-hani, M. N. Marsono, and R. Bakhteri, “Biometric encryption based on a
fuzzy vault scheme with a fast chaff generation algorithm,” Futur. Gener. Comput.
Syst., vol. 29, no. 3, pp. 800–810, 2013.
[45] J. Bringer, “Fuzzy Vault for Multiple Users,” pp. 67–81, 2012.
[46] K. Y. Moon and D. Moon, “Biometrics Information Protection using Fuzzy Vault
Scheme,” 2012.
[47] C. Rathgeb, B. Tams, J. Wagner, and C. Busch, “Unlinkable improved multi-biometric
iris fuzzy vault,” EURASIP J. Inf. Secur., 2016.
[48] G. Revadigar, S. Member, C. Javali, and S. Member, “Accelerometer and Fuzzy Vault-
 Based Secure Group Key Generation and Sharing Protocol for Smart Wearables,” vol.
12, no. 10, pp. 2467–2482, 2017.
[49] V. Sujitha and D. Chitra, “Highly secure palmprint based biometric template using
fuzzy vault,” no. February, pp. 1–7, 2018.
[50] K. Koptyra and M. R. Ogiela, “Multiply information coding and hiding using fuzzy
vault,” Soft Comput., vol. 23, no. 12, pp. 4357–4366, 2019.
[51] D. Prajyoti, “Securing ATM Pins and Passwords Using Fingerprint Based Fuzzy Vault
System,” 2018.
[52] L. You, T. Wang, T. Wang, and L. You, “A novel fuzzy vault scheme based on
fingerprint and finger vein feature fusion,” Soft Comput., vol. 23, no. 11, pp. 3843–
3851, 2019.
[53] V. Sujitha and D. Chitra, “A Novel Technique for Multi Biometric Cryptosystem
Using Fuzzy Vault,” 2019.
[54] W. Ponce-hernandez, R. Blanco-gonzalo, J. Liu-jimenez, and R. Sanchez-reillo,
“Template protection approaches : Fuzzy Vault scheme,” 2019.
[55] W. Ponce-hernandez, R. Sanchez-reillo, R. Blanco-gonzalo, and J. Liu-jimenez,
“Fuzzy Vault Scheme Based on Fixed-Length Templates Applied to Dynamic
Signature Verification,” vol. 8, 2020.
[56] X. Dong, S. Kim, Z. H. E. Jin, J. Y. Hwang, S. Cho, and A. J. I. N. Teoh, “Secure
Chaff-less Fuzzy Vault for Face Identification,” vol. 17, no. 3, 2021.
[57] V. Singh, B. Surya, and P. Ity, “An enhanced fuzzy vault to secure the fingerprint
templates,” Multimed. Tools Appl., no. July, pp. 33055–33073, 2021.
[58] “Homomorphic Encryption,” pp. 27–46, 2014.
[59] W. Abel and A. Torres, “Effectiveness of Fully Homomorphic Encryption to Preserve
the Privacy of Biometric Data,” 2014.
[60] M. Gomez-barrero et al., “Implementation of Fixed-Length Template Protection Based
on Homomorphic Encryption with Application to Signature Biometrics,” pp. 191–198.
[61] M. Gomez-barrero, E. Maiorana, J. Galbally, P. Campisi, and J. Fierrez, “Multi-
biometric template protection based on Homomorphic Encryption,” Pattern Recognit.,
vol. 67, pp. 149–163, 2017.
[62] A. Nautsch, S. Isadskiy, J. Kolberg, M. Gomez-barrero, and C. Busch, “Homomorphic
Encryption for Speaker Recognition : Protection of Biometric Templates and Vendor
Model Parameters.”
[63] P. Drozdowski, N. Buchmann, C. Rathgeb, M. Margraf, and C. Busch, “On the
Application of Homomorphic Encryption to Face Identification,” 2019.
[64] M. Kumar, M. V. N. K. Prasad, M. Verma, and U. S. N. Raju, “Secure and verifiable
iris authentication system using fully homomorphic encryption,” Comput. Electr. Eng.,
vol. 89, no. August 2020, p. 106924, 2021.
[65] E. Okoh, “Template protection in biometric-based mobile scenarios,” pp. 1–2, 2021.
[66] H. Tamiya, “Improved Post-quantum-secure Face Template Protection System Based
 on Packed Homomorphic Encryption,” pp. 1–5, 2021.
[67] M. Barni et al., “A Privacy-compliant Fingerprint Recognition System Based on
Homomorphic Encryption and Fingercode Templates.”
[68] A. Bassit, F. Hahn, C. Zeinstra, R. Veldhuis, and A. Peter, “Bloom Filter vs
Homomorphic Encryption : Which approach protects the biometric data and satisfies
ISO / IEC 24745 ?,” 2021.
[69] X. Yang, H. Zhu, S. Zhang, R. Lu, and X. Gao, “An Efficient and Privacy-Preserving
Biometric Identification Scheme Based on the FITing-Tree,” vol. 2021, 2021.