The Role of Biometric in Banking: A Review

Abstract

Biometrics plays a pivotal role in enhancing security, ensuring accurate identification, and offering convenient solutions across diverse industries. Its uniqueness, reliability, and potential for future advancements establish it as a crucial and valuable field in today's digital landscape. Fingerprint authentication in ATMs presents primary advantages such as heightened security through distinctive identification and user convenience by eliminating the reliance on PINs or passwords. This research paper focuses on conducting a comprehensive review and comparative analysis of various approaches for fingerprint identification, aiming to contribute to the understanding of effective and efficient methods in the context of ATM authentication.

Full text

EAI Endorsed Transactions
on AI and Robotics Review Article
The Role of Biometric in Banking: A Review
Mehdi Marani1,∗, Morteza Soltani2, Mina Bahadori3, Masoumeh Soleimani4, Mehdi Davari5,
Atajahangir Moshayedi6
1Department of Mechanical Engineering, Khomeinishahr Branch, Islamic Azad University, Isfahan, Iran
2Department of Industrial Engineering, Clemson University, SC, USA
3Department of Industrial Engineering, Clemson University, SC, USA
4Department of Mathematics and Statistical Sciences, Clemson University, SC, USA
5Department of Management, Isfahan Branch, Islamic Azad University, Isfahan, Iran
6School of Information Engineering, Jiangxi University of Science and Technology, No 86, Hong qi Ave, Ganzhou,
Jiangxi, 341000, China
Abstract
Biometrics plays a pivotal role in enhancing security, ensuring accurate identification, and offering convenient
solutions across diverse industries. Its uniqueness, reliability, and potential for future advancements establish
it as a crucial and valuable field in today’s digital landscape. Fingerprint authentication in ATMs presents
primary advantages such as heightened security through distinctive identification and user convenience by
eliminating the reliance on PINs or passwords. This research paper focuses on conducting a comprehensive
review and comparative analysis of various approaches for fingerprint identification, aiming to contribute to
the understanding of effective and efficient methods in the context of ATM authentication.
Received on 03 August 2023; accepted on 18 August 2023; published on 21 August 2023
Keywords: Fingerprint Sensor, Microcontroller, Identification Methods, Hardware
Copyright © 2023 Marani et al., licensed to EAI. This is an open access article distributed under the terms of the
Creative Commons Attribution license (http://creativecommons.org/licenses/by/4.0/), which permits unlimited
use, distribution and reproduction in any medium so long as the original work is properly cited.
doi:10.4108/airo.3676
1. Introduction
Biometrics refers to the measurement and analysis of
distinct physical or behavioral attributes unique to indi-
viduals, a concept that has gained prominence across
various sectors. Notably, the integration of biometrics in
the banking industry has emerged as a pivotal area of
exploration. Biometric identification presents an inno-
vative approach compared to traditional methods like
passwords or identification cards, as biometric traits
are immune to issues such as theft, replication, pur-
chase, or forgery. This heightened level of dependability
has spurred considerable interest in adopting biomet-
ric solutions within the banking domain, reflecting a
notable step forward in bolstering security and reliabil-
ity.
Numerous research studies have been undertaken
to delve into the potential applications of biometrics
∗Corresponding author. Email: msoleim@g.clemson.edu
within the banking domain, and this paper offers
a compilation of selected investigations. A specific
area of focus involves the exploration of fingerprint-
based Automated Teller Machines (ATMs), an arena
riddled with an array of intricate challenges. These
challenges span a wide spectrum, encompassing
factors like financial feasibility, concerns regarding
device portability, the establishment of robust security
measures, the acceptance of the technology among
users, the efficiency of enrollment procedures, the
resilience of the system, as well as issues related to
privacy and defect rates.
However, the primary objective of this manuscript
is to meticulously tackle the initial three challenges:
namely, cost, portability, and security. The research
endeavors outlined in this paper are meticulously
designed to provide an all-encompassing evaluation
and comparative analysis of various methodologies
implemented in this sphere. The ultimate aim is to
not only enhance the cost-efficiency, mobility, and
1
EAI Endorsed Transactions on
AI and Robotics

security aspects of fingerprint-based ATMs but also
to contribute significantly to elevating user experience
and the overall operational effectiveness of the banking
landscape.
By thoroughly examining the existing literature and
synthesizing the information, the study identifies key
research gaps and areas that warrant further exami-
nation. This research seeks to explore advancements
and challenges in biometric authentication methods,
specifically within the context of fingerprint identifi-
cation devices employed in ATMs. The primary aim is
to enhance the existing pool of knowledge and offer
valuable input for the progress of the field in the future.
1.1. Research Contribution
1. Reviewing different approaches to fingerprint
identification.
2. comparing the results of these methods to provide
valuable insights for future researchers and contribute
to the advancement of biometric authentication systems
in ATM technology.
2. Review
T. Sabhanayagam et al. (2018) Offer a comprehensive
overview of diverse biometric systems and their func-
tional applications. Furthermore, the study concludes
that while biometric recognition systems effectively
address the limitations of traditional methods, it is
important to consider potential challenges arising from
the continuous evolution of biometric technology [1].
Khan et al. (2020) Involved in a separate endeavor that
aimed to comprehend health data derived from wear-
able IoT technology. The team conducted an extensive
review, discussion, and offered recommendations on
utilizing this data as a biometric for computer security
purposes. Several limitations were identified, including
the ever-changing characteristics of the human body,
sensor aging and deterioration, difficulties in selecting
an appropriate population for biometric system assess-
ment, sensor cost, effects on accuracy, lack of standard-
ized techniques, and the inadequate security measures
prevalent in biometric systems [2]. M. Gayathri et al.
(2020) Conduct a comprehensive survey on the current
state of biometric technology, covering various aspects
such as different types of biometric traits, the tech-
niques employed for extracting these traits’ features,
and the diverse application areas where various bio-
metric traits find utility. The authors emphasize the
paramount importance of safeguarding biometric data
due to its sensitive nature. Despite biometric tech-
nology significantly reducing the problems associated
with traditional forms of theft, it is acknowledged that
certain vulnerabilities still exist. Hackers have demon-
strated the ability to breach biometric data, under-
scoring the critical need for robust biometric template
protection measures[3]. To illustrate the current trends
in biometrics, the authors cite specific examples. These
include the utilization of ECG biometrics control in cars
for monitoring the health of drivers and passengers,
as well as the adoption of palm recognition in smart
panda buses Alsaadi (2021) summarized the existing
behavioral biometrics systems and offers an exploration
of the key advantages and disadvantages associated
with popular behavioral biometrics technologies [4]. In
a review paper by Yang et al. (2021), The researchers
delve into a diverse array of strategies aimed at mitigat-
ing vulnerabilities in various layers of the Internet of
Things (IoT) architecture. The authors underscore that
every biometric trait, whether employed individually
or in combination, exhibits unique strengths as well as
notable limitations [5].
Biometric traits, such as fingerprints, iris patterns,
and facial features, are distinctive to each individ-
ual. Among these biometric features, fingerprints have
held the top position as the benchmark for individual
identification for many years. In the 1990s, fingerprint
scanning using optical, ultrasonic, and infrared imag-
ing techniques was introduced, and measurements of
pressure, temperature, and electric capacitance were
used to detect patterns on the surface of the finger and
convert them into electrical signals. The efficiency of
a fingerprint recognition system is contingent on the
accuracy of the algorithms used for feature extraction.
Today, there are multiple approaches with acceptable
results in fingerprint feature extraction. The problem
arises when traditional methods fail to analyze finger-
print textures under low-quality conditions. Thus, var-
ious methods, including neural networks, traditional
methods, and hybrid approaches, have been employed
to extract features and match them with samples in the
database [6].
Optimization algorithms play a significant role in
enhancing the performance and effectiveness of biomet-
ric systems. Here, we introduce some papers that focus
on optimization algorithms: Ansaripour et al. applied
neural networks and optimization techniques to miti-
gate welding challenges and this approach can be anal-
ogous to improving biometric system performance and
security [7]. Daniali et al. optimize a shell and tube heat
exchanger for thermal efficiency and total cost using
Copper oxide/Iranol refinery oil nanofluids. Employing
natural base inspired algorithms, it generates a Pareto
frontier curve, demonstrating the trade-offbetween
efficiency and cost considerations [8]. Babajamali et al.
explore multi-objective optimization of tandem cold
rolling parameters using NSGA-II and Pareto-optimal
front, aiming to enhance reductions and inter-stand
tensions. The obtained optimized rolling schedules
demonstrate improved performance, echoing biometric
systems’ pursuit of better accuracy and efficiency [9].
Barnoon et al. analyze proton exchange membrane fuel
2
Marani et al.
EAI Endorsed Transactions on
AI and Robotics

cell (PEMFC) cooling, stress, and displacement under
various conditions, utilizing multi-objective optimiza-
tion to determine optimal plate thickness and number
for temperature, stress, and displacement reduction.
Similar to optimizing biometric systems, the research
aims to enhance performance and longevity while con-
sidering factors like temperature control and stress
management [10]. Khan et al. introduce a gait recogni-
tion framework employing deep learning and Bayesian
optimization, analogous to biometric identification. The
framework combines motion region extraction, hyper-
parameter optimization, and feature fusion, achieving
high accuracy on public datasets similar to biometric
systems’ accuracy goals [11]. M Soltani presents a model
for optimizing spare parts supply chains and condition-
based maintenance, akin to the integration of com-
ponents in biometric systems. This paper introduces
innovative policies for spare part ordering based on
system deterioration and optimizes decision variables
for improved availability and cost efficiency, mirroring
concerns in both logistics and biometric fields [12].
Soleimani et al. apply imbalanced data in medical
classification, resembling the challenge of recognizing
underrepresented traits in biometric systems. they pro-
pose a feature selection method using a support vector
machine and wrapper approach, optimizing accuracy to
%99.6 through neural network optimization, mirroring
biometric systems’ quest for accuracy in diverse trait
recognition [13].
In their research, Sun et al. address presentation
attack detection (PAD) in fingerprint recognition sys-
tems, reflecting the challenge of distinguishing genuine
users from fraudulent attempts in biometric systems.
They introduce a novel method utilizing optical coher-
ence tomography (OCT) features, achieving accurate
attack detection with a %4 Equal Error Rate (EER),
mirroring biometric systems’ emphasis on precise iden-
tification [14].
Nowadays, considering the increasing need for secu-
rity, this technology is being utilized in various appli-
cations such as attendance management, secure locks,
ATMs, card reader devices, control and surveillance
of agricultural machinery, vehicles, banking security
system, and so on. Furthermore, to achieve each of the
above objectives, it is necessary to utilize suitable hard-
ware along with user-friendly programming specifically
designed for it. This ensures complete monitoring and
control of user equipment while maintaining security
and minimizing the required time. It is worth mention-
ing that the hardware used in these systems should pos-
sess certain feathers such as simplicity, compactness,
affordability, portability, ease of use, and, ideally be
adaptable for various applications. To provide further
illustration, Kumar et al. have developed a biometric-
based security lock system utilizing fingerprint recog-
nition, showcasing one of the practical applications of
fingerprint technology [15]. This design incorporates
a two-step identity verification process, consisting of
credential authentication and fingerprint verification.
Additionally, it has the capability to capture images of
unauthorized users. The fingerprint is utilized as an
identity verification system in this design. The opera-
tional steps of this design are, to enter your password
using the keypad, and place your fingerprint on the
fingerprint scanner. If the fingerprint is unauthorized,
the image will be captured by the camera module and
stored in the computer system, If the password and
fingerprint belong to an authorized person, the entry
door connected to the DC motor will be opened, now
you can access your shelf. In this project, two micro-
controllers, Atmega 16 and Pic16F877A, are employed.
Both microcontrollers communicate with each other
through various ports. Atmega 16 is connected to
peripheral devices such as an LCD, keypad, and fin-
gerprint module. Pic16F877A is connected to the DC
motor, buzzer, and computer system. The fingerprint
sensor used in this project is an optical type, and it scans
the fingerprint using light-sensitive diodes, storing the
signals as bright and dark pixels. An analog-to-digital
converter is present in the scanner to convert the analog
signals into digital form. This research indicates that
designing a flawless security shelf can be an effective
solution for addressing the significant problem of theft
in today’s world. KM and et al. have designed a security
system based on fingerprint and RFID sensors. This sys-
tem adopts a two-factor authentication method, leading
to improved safety and effectiveness within the system.
In this design, the RFID reader first reads a ten-digit
code from the corresponding tag. These tags or labels
contain microprocessors that store the ID or identifier
of each object. It should be noted that when radio
waves reach the RFID antenna, they create a magnetic
field. The tags draw power from this magnetic field and
become capable of transmitting information. The fin-
gerprint sensor, on the other hand, compares the input
image with the registered data and sends a verification
signal to the computer. If both the RFID and fingerprint
verification match, the microcontroller commands the
motor associated with the doors, and they open by the
motor’s rotation. The hardware components used in this
design include a pic microcontroller, fingerprint sensor,
computer, RFID tags, electronic relay, and DC motor.
The standard employed in the tags is based on the
magnetic fields generated at close and far distances. LF
(Low Frequency) and HF (High Frequency) frequencies
are used for close distances, while UHF (Ultra-High
Frequency) or microwave frequencies are used for far
distances [16]. Figure 1 illustrates the tags and magnetic
fields at far and close distances, respectively.
The researchers’ findings point out that by develop-
ing and implementing a suitable and efficient security
system in banks, it can foster a sense of trust among
3
The Role of Biometric in Banking: A Review
EAI Endorsed Transactions on
AI and Robotics

Figure 1. The magnetic fields created at close and
distant distances
users, encouraging them to entrust their assets to the
banking institutions. Gill KR, et al. [17] have designed
a system for starting a car using a fingerprint. The
first companies to adopt this system for vehicles were
Mercedes-Benz and Volkswagen. In this design, the
process begins with sampling the owner’s fingerprint.
The main identity verification core of the car’s switch
is a microcontroller. The fingerprint sensor captures
the user’s fingerprint and sends a signal to the micro-
controller. The microcontroller performs a comparison
between the scanned fingerprint and the one stored in
its database. Once a successful match is detected, the
car is activated and powered on. In this setup, a GSM
module can also be used to fulfill an important role.
Whenever an unauthorized user attempts to scan their
fingerprint, a message is sent to all authorized users.
Since the required power supply is 5 volts, while the
car battery is 12 volts, an IC 7805 is used to obtain
a 5-volt voltage. The hardware components used in
this design include: The implemented setup includes
the AT89S52 microcontroller, power supply, fingerprint
sensor (R305), display, car ignition system, and GSM
modem (SIM900A). The R305 fingerprint sensor is cho-
sen for its low power consumption, affordability, supe-
rior performance compared to similar sensors, and easy
implementation. Similarly, the SIM900A GSM module
is a cost-effective and user-friendly module with mes-
saging capabilities and, when equipped with a SIM
card and internet connection, can also send emails.
The design proposes the possibility of incorporating
iris scanning or heartbeat detection as alternative iden-
tification methods in case of incomplete fingerprints
or finger-related issues in the future. The researchers’
findings indicate that the car ignition system using a
fingerprint sensor leads to savings in traditional keys
for starting the vehicle and can also be used when the
user forgets to bring the car key, as the car can be
turned on by authenticating the user’s fingerprint using
the sensor. Madhu R and others [18] have designed a
method for remote communication with a device using
fingerprint identification along with GSM and GPS
capabilities. The software implementation process is as
follows: initially, the user needs to place their finger
on the fingerprint module. Then, the microcontroller
compares the fingerprint image with its database and,
if valid, sends a message to the device owner requesting
access. If the device owner sends an access grant mes-
sage, The end-user possesses the capability to control
all the devices simultaneously. Table 1 presents a com-
prehensive comparison between the proposed design
and other existing designs. The hardware components
employed in this system comprise a GSM module, GPS
module, and Arduino board [19], fingerprint sensor,
display, and electronic relay with a junction box. The
research indicates that this design enables multiple
individuals to have access and control over the device’s
operation without necessarily being physically close
to the device. The increased safety and flexibility for
various applications, including agricultural machinery
control, vehicle control, and theft detection of vehicles,
are among the results of this design. In their 2016
paper, Singh et al. detailed the procedures of finger-
print verification and explored strategies to enhance
the performance of the fingerprint biometric system,
thus facilitating smooth access to the system. It resulted
in higher security, and saves time, and solves several
issues related to the input system [20]. In a 2018
study conducted by Manzoor, an in-depth analysis of
various fingerprint biometric systems was presented,
along with a detailed explanation of a fingerprint-
based biometric system. The research demonstrated a
trade-offbetween the security of biometric information
and the overall system performance. Manzoor empha-
sized that achieving a secured biometric system with
both lower equal error rates and higher identification
accuracy continues to pose a significant challenge for
researchers[21]. Sagayam et al. (2019) identify and clas-
sify fingerprint recognition by using the Euclidean dis-
tance and neural network classifier (NNC). This method
improves the performance in some aspects like: time
and accuracy. By Euclidean distance and NNC one can
analyze the test image and the input image accuracy,
but when the image is enrolled in different angles,
the matching between the input image and the test
image is not straightforward [22]. In their 2020 study,
Ishak et al. designed software specifically catering to
high-ranking officers in a security company, offering a
range of ready-to-use formats. To ensure data security,
all information is securely stored in a database. The
system includes a unique fingerprint authentication
feature for user access. Integrating the Secure Biometric
Lock System for files and applications brings about
several advantages, such as saving time and resources,
minimizing paperwork, preventing errors, facilitating
effective communication with senior staffor employees,
ensuring proper etiquette, and providing other valuable
benefits [23]. In their 2021 research, Marco Ferretti et
4
Marani et al.
EAI Endorsed Transactions on
AI and Robotics

al. present a distributed solution for behavioral IoT fin-
gerprinting. They recognize the scalability challenges
posed by updating fingerprint models for device config-
uration variations in centralized solutions. To address
this, they strategically target specific nodes, includ-
ing gateways, as part of their distributed approach.
Within the IoT ecosystem, a novel framework called
H2O (Human to Object) has been proposed, designed
to enable object monitoring through trained classifi-
cation models, ensuring a scalable solution. Dedicated
controller nodes integrated into the Internet Service
Provider (ISP) efficiently conduct the training process.
H2O introduces a mechanism to authenticate the iden-
tity of objects or humans based on their claims. The
study includes a security analysis of the framework,
which encompasses an attack model, demonstrating
that the proposed solution effectively achieves its objec-
tives even when faced with potential attacks [26]. In
their latest research, Yin et al. (2021) introduce an
IoT-oriented privacy-preserving fingerprint authentica-
tion system. This system incorporates four essential
components: minutiae extraction, a cancelable binary
template based on the minutia cylinder-code (MCC)
generated using a novel normalized random projec-
tion technique, a lightweight, privacy-preserving tem-
plate constructed through pairwise Boolean operations,
and fingerprint matching. Notably, this groundbreak-
ing system represents the first privacy-preserving, can-
celable fingerprint authentication solution specifically
designed to meet the demands of resource-constrained
IoT environments [27]. In their 2022 study, Nontha-
putha et al. explore an innovative IoT-based solution
for a smart biometric fingerprint circular key storage
cabinet. This advanced system empowers students to
conveniently borrow and return machinery keys via
a user-friendly touch screen and biometric fingerprint
scan. All key transactions are efficiently recorded in the
database, ensuring seamless tracking and management.
The researchers have developed a user-centric software
interface, ensuring students can effortlessly access the
keys they need at any time, making the borrowing
and returning process convenient and straightforward
[28]. In their 2023 study, Siregar et al. endeavor to
design a fingerprint sensor implementation system for
a fingerprint reader prototype, incorporating a micro-
controller. The central goal of this system is to achieve
rapid and accurate attendance recording while ensur-
ing its robustness against tampering and manipulation
[29]. Lastra, et al. [30] conducted extensive research
on efficient fingerprint identification using graphics
processors (GPUs). Fingerprint recognition is widely
utilized in various biometric identification systems.
However, the fingerprint-matching process poses con-
siderable computational challenges, requiring stream-
lined processing methods. With potentially large finger-
print databases, the scalability of models relies on the
number of fingerprints and points in each fingerprint.
To cater to the need for accelerated processing, the
researchers put forth a unique fingerprint matching
algorithm centered around minutiae, specially opti-
mized for GPU-based massively parallel processing.
The study’s results highlight the minutiae-based fin-
gerprint matching algorithm as one of the leading
methods among biometric identification approaches.
Hambalık et al. [31] conducted a research study focused
on fingerprint recognition systems using Artificial Neu-
ral Networks (ANNs). Their investigation explored the
potential integration of ANNs as feature extractors
within the fingerprint recognition process. As a result of
their investigation, the researchers created a complete
and operational software system designed to perform
fingerprint recognition, consisting of modules for high-
resolution fingerprint sensors, image enhancement, fea-
ture extraction, and fingerprint matching. The study’s
results demonstrated the significant impact of neural
networks on the overall detection rate, particularly in
scenarios involving low-quality fingerprint images.
In recent times, certain banks have initiated trials of
biometric ATM systems in specific regions or limited
pilot programs. These initiatives require customers to
enroll their biometric data (e.g., fingerprints) with the
bank and utilize biometric authentication at designated
ATMs. Biometric data is securely stored and utilized
for subsequent ATM transactions, eliminating the need
for physical cards or PINs. However, it’s essential to
acknowledge that the adoption and implementation
of biometric ATMs may vary based on the country,
financial institution, and regulatory requirements.
While biometric authentication offers advantages in
terms of security and convenience, it also brings
challenges such as privacy concerns, technological
integration, and cost implications [32].
In the realm of ATM security enhancements, various
works have been explored, and some will be reviewed
here. R. Murugesh (2012) introduced an innovative
approach to bolster ATM security by replacing
traditional ATM cards with fingerprints. Utilizing the
AES 256 algorithm for PIN and OTP encryption,
the study found that transitioning to a biometric
system could streamline transactions, ensuring ease,
reliability, and a stress-free experience without the need
for physical cards. The robust AES 256 encryption
provided solid security features, and the inclusion
of a steganography mechanism further fortified the
system against middleman attacks. With cost-effective
biometric scanners readily available, this system
promises users a seamless and secure experience,
striking a perfect balance between convenience and
safeguarding data integrity [33]. In their 2017 study,
Taralekar et al. introduced an innovative "One touch
multi-banking transaction system using biometric and
GSM authentication" for ATM terminal customer
5
The Role of Biometric in Banking: A Review
EAI Endorsed Transactions on
AI and Robotics

recognition. The proposed system effectively addresses
the shortcomings of the traditional approach by
introducing upgraded security features for seamless
and secure transactions [34].
In their 2018 research, B. Saranraj et al. proposed
a highly secure authentication system for ATMs. This
system ensures that only the valid cardholder can access
the ATM, and entering the ATM center requires both
the account holder’s ATM card and their knowledge. In
case an unauthorized person tries to use the ATM card,
the system sends an OTP to the account holder, and
only upon entering this OTP into the ATM machine, the
user is allowed to proceed with cash withdrawal. This
advanced approach offers excellent efficiency while
effectively preventing any illicit transactions [35]. In
their 2019 study, Alzamel et al. proposed an innovative
integrated fingerprint biometric authentication system
designed specifically for securing the Point of Sale
(POS) network. This new security option complements
the use of ATM cards. The research findings revealed
that this service offers customers a secure alternative
for conducting daily transactions, reducing the reliance
solely on the ATM card. The questionnaire results
indicated a high demand among customers for a
safer substitute for the card, indicating strong backing
for the proposal, as it has the potential to mitigate
issues associated with traditional ATM cards [36]. In
2019, Bataev published a paper aimed at evaluating
the economic efficiency of specific devices. The study
employed the Total Cost of Ownership (TCO) method,
incorporating expert estimates of stolen funds in the
Russian Federation to conduct the calculations. The
findings demonstrated significant economic efficiency
of these ATMs; however, they also highlighted the
considerable expenses involved in implementing such
ATMs within the banking systems of individual banks
[37]. In 2022, T. Sangeetha et al. conducted research
with a specific focus on incorporating a fingerprint-
based method into the ATM system to bolster its
security. The primary objective was to develop a
more resilient security system using fingerprint-based
ATMs. Their proposed system involved converting
fingerprints into unique string values, which were then
securely stored in a vast cloud memory within the EC2
database. During a transaction, the user’s fingerprint
was matched with the unique string in the cloud to
facilitate authentication [38].
The ATM system based on fingerprint authentication
was split into two stages: enrollment and authenti-
cation. In the enrollment phase, users were required
to register up to 3-4 fingers to establish a protec-
tion threshold and accommodate various finger-related
issues, such as wet, dry, skewed, dirty, cut, or worn
fingers. The enrollment process ensured the system’s
flexibility and adaptability to real-life scenarios [39].
In the authentication phase, users could effortlessly
conduct transactions by placing their finger on
the biometric scanner. The fingerprint scan was
then compared against the database containing all
authenticated user fingerprints, ensuring a swift and
secure transaction process. This novel approach aimed
to strengthen the ATM system’s security and provide
users with a seamless experience through biometric
authentication [40]. In their 2021 research, Hochwarter
et al. conducted a study to examine the general attitudes
and acceptance of biometrics within the Austrian
public. The data was obtained through an online
survey, revealing that while a notable minority showed
resistance to the concept of ATMs with biometrics, a
larger segment expressed a positive inclination towards
such a system. However, the study did not indicate a
strong preference for any specific biometric approach
for ATMs among the Austrian populace [41].
Moshayedi et al, (2023) [42] proposed a novel sys-
tem design called PFIB (Fingerprint-based ATM). They
found The classification accuracy rate (CAR) is calcu-
lated to be 95%. This indicates that the fingerprint
recognition system achieves a high level of accuracy,
correctly classifying 95% of the samples. The true pos-
itive rate (TPR) is computed as 93.3%. By this method,
ATMs can overcome challenges related to maintenance,
cost, security, and efficiency, improving overall perfor-
mance, and reducing operational expenses. Addition-
ally, The design allows simultaneous or independent
use of the card reader, potential cost savings by replac-
ing existing card readers, a significant reduction in out-
of-service ATMs, portability and compact size, versatil-
ity for various purposes, low manufacturing cost, high
transmission speed, and enhanced security measures.
Upon reviewing these papers, it becomes evident that
the majority of fingerprint recognition devices share
several common components, as highlighted in Section
2.
3. The structure of a fingerprint recognition device
A cutting-edge fingerprint recognition device harmo-
nizes hardware and software for meticulous identifica-
tion. Its components, showcased in Figure 2, include
fingerprint sensors capturing intricate details, while
communication protocols ensure seamless data trans-
mission. Processors analyze data, unveiling patterns,
and the identification module employs algorithms for
precise matches, echoed by the display module’s out-
comes. This symphony of elements defines the fin-
gerprint recognition landscape’s sophistication.In this
harmonious integration, the device not only exemplifies
technical finesse but also sets a benchmark for reliable
and secure identity verification.
6
Marani et al.
EAI Endorsed Transactions on
AI and Robotics

Figure 2. The design of a fingerprint recognition device
3.1. Sensor
Fingerprint registration devices are divided into two
models based on the method of fingerprint acquisition:
live scan and offline scan. In offline models, the process
involves inking the fingers and pressing them onto
paper to obtain fingerprints. Afterward, the paper is
either captured or scanned to acquire the fingerprint
images.
On the other hand, live scan methods work by
digitally capturing the fingerprint when it touches
the sensors. The primary use of offline methods is
for identifying the fingerprints of criminals at the
crime scene. Digital scanners are categorized based on
resolution, pixel count, sensor area, accuracy, and other
factors. Live scan digital scanners are divided into five
categories: optical, capacitive, thermal, pressure-based,
and ultrasonic.
There is a widespread application of fingerprint
recognition sensors, they are used for identity confir-
mation in self-service kiosks [43], organizations, gov-
ernment agencies, universities, and educational institu-
tions for attendance tracking [44], security systems for
doors, vehicles, warehouses, and so on. In the following,
we review some of these applications, and then we
introduce some common types of sensors.
The most common types of fingerprint sensors used
include optical, capacitive, and ultrasonic sensors we
present a brief explanation about them here.
(i) Optical Sensors: This method is based on cap-
turing an image and detecting specific patterns
within it. This technology uses a series of default
algorithms to examine the dark and light points
present in the image, identifying the ridges and
valleys on the skin. Optical sensors consist of a
multitude of diodes, with some of them respon-
sible for providing the necessary light for captur-
ing the fingerprint. The advantages of these sen-
sors include their simple manufacturing technol-
ogy. However, they occupy a significant amount
of memory, can only record two-dimensional
images, and have lower security and higher pro-
duction costs.
(ii) Capacitive Sensors: This technology is founded
on an electronic circuit and an interconnected
network of capacitors, forming the basis of its
functionality. Instead of capturing an image,
capacitive sensors use an array of capacitive
circuits to record information related to the users’
fingerprints. Since capacitors can store electric
charges, their connection to conductive plates on
the sensor surface allows for precise information
retrieval regarding the ridges’ placement and the
number of stored charges within the capacitors.
These sensors have higher security, and it is not
possible to misuse the information from another
fake image. However, due to different materials
causing variations in the charge levels of the
capacitors, the likelihood of compromising the
system’s security with other materials is almost
negligible. Nonetheless, they are susceptible to
software and hardware hacking.
(iii) Ultrasonic Sensors: This type of sensor utilizes a
transmitter and receiver of ultrasonic waves to
capture fingerprint information. When the finger
is placed on the sensor, a high-frequency sound
pulse is directed toward it. The skin absorbs
certain segments of this pulse, while it reflects the
remaining components. Based on the variations
in the reflected waves due to the protrusions and
depressions on the skin surface, the receiver of the
sensor can obtain precise information about the
7
The Role of Biometric in Banking: A Review
EAI Endorsed Transactions on
AI and Robotics

users’ fingerprints by examining the intensity of
the pulse at different points. These sensors have
the capability to record three-dimensional images
and provide high-level security. However, they
have a more complex manufacturing technology
compared to the previous two models.
Table 1provides a comparison of these three sensor
types.
Table 1. Comparison of Fingerprint Sensor Types
Sensor
Type
Advantages Disadvantages
(i) Simple man-
ufacturing
technology
-excessive memory
-Low cost production
-Just two dimensional images
-Higher error rates
-Low security
(ii) Higher
security
-Susceptible to software and
hardware hacking
(iii) recording
three
dimensional
-Higher
security
-More complex manufactur-
ing technology images
3.2. Communication Protocol
To connect and establish communication between the
processor and the sensor, different protocols are used.
The most common ones are I2C, SPI, and UART, which
are all part of the serial communication set. Now, we
introduce them as follows and then these protocols are
compared to each other in Table 2.
The I2C protocol is a combination of the finest
attributes of SPI and UART. With I2C, it is possible
to connect multiple slaves to one master (like SPI) or
use multiple masters to control one or more slaves.
This feature becomes especially valuable when utilizing
multiple microcontrollers to transmit data to a memory
card or display it on an LCD. Components such as old
displays, barometric pressure sensors, gyroscopes, and
accelerometers use this protocol.
SPI (Serial Peripheral Interface) is a widely used
communication protocol found in various modules,
including SD card units, RFID card reader units, and
2.4 GHz wireless transceiver modules. Its primary
function is to enable seamless data transfer, allowing
continuous transmission or reception of any number
of bits. Communication occurs in a master-slave
relationship, with the master component (usually
a microcontroller) sending commands to the slave
component (such as a sensor, display, or Memory
circuit). While the simplest SPI setup involves one
master and one slave, it is also possible for one master
to control multiple slaves.
Figure 3. Fingerprint patterns
In contrast, UART communication involves a direct
connection between two devices. The transmitting
device receives parallel data from a control unit,
such as a CPU, and converts it into a serial format
before sending it to the receiving device. The receiving
device subsequently converts the serial data back to
parallel format, and the data is transmitted from the
transmitting device’s TX pin to the receiving device’s
RX pin.
For an extensive analysis and comparison of Com-
munication Protocols, please refer to the accompanying
Table 2.
3.3. Identification methods
In this section we explain the underlying principles
and mechanisms of fingerprint recognition, firstly, then
we review some of the methods used for fingerprint
authorization and at the end we will compare various
methods used in fingerprint recognition. a dynamic
pattern of ridges and valleys characterizes the skin on
the palm and sole of the foot. The whole extent of
the palm and fingers is covered with continuous, thin
lines, known as friction ridges. These friction ridges
play a crucial role in increasing friction with objects,
providing a stronger grip, and enhancing the sense
of touch when in contact with various surfaces. In
addition to these functions, the identity of individuals
can be determined through these friction ridges. This is
because these ridges are unique and unchangeable for
everyone’s fingers. Locations, where the friction ridges
are suddenly interrupted or divided into two or more
branches, are called minutiae. In general, the main
fingerprint patterns are divided into three categories:
arches, rings, and spirals, as shown in Figure 3, [45].
By analyzing these patterns, characteristics such as
minutiae can be extracted.
By using local minutiae structures, one can quickly
find relative matches between two fingerprint samples,
8
Marani et al.
EAI Endorsed Transactions on
AI and Robotics

Table 2. Comparison of Communication Protocols
Communication Protocol Advantages Disadvantages
UART •Use of only two wires
•No need for a separate clock signal
•Existence of parity bit for error checking
•Data packet structure can be adjusted depending
on the preferences set at both terminations
•Availability of extensive documentation and
widespread implementation methods
•The size of transmitted data is limited to a
maximum of 9 bits
•Absence of support for multiple masters (con-
trollers) and multiple slaves (controlled devices)
in the system
•The baud rates of both sides can have a maximum
variance of 10% from each other
SPI •There are no start or stop bits, facilitating the
continuous transfer of data
•There is no sophisticated addressing mechanism
for controlled devices, similar to the one present
in I2C
•Faster data transmission rate in comparison to
I2C (approximately double the speed)
•With separate lines for MISO and MOSI,
simultaneous data reception and transmission are
possible
•The use of four wires (compared to I2C and
UART, which use two wires)
•Absence of explicit indication for confirming
the correct reception of data (unlike I2C, which
possesses this capability)
•Lack of error detection mechanism such as parity
in UART
•The presence of only one master is the only
option provided
I2C •Use of two wires
•Support for multiple masters and multiple slaves
•Capability to confirm or decline data transfer
using ACK/NACK bits
•Less complex hardware configuration in compar-
ison to the UART approach •Familiar and exten-
sively utilized protocol.
•Reduced data transfer speed in comparison to the
SPI technique
•Data length is restricted to a maximum of 8 bits
•Hardware implementation may pose more com-
plexities when compared to the SPI method
ensuring a complete match between them. Represent-
ing the minutiae as nodes and connecting them with
edges, based on their proximity or spatial relation-
ships allows the construction of a graph for each fin-
gerprint sample. Various graph-based algorithms and
techniques from graph theory can then be utilized to
measure the similarity or dissimilarity between the two
graphs, indicating the degree of match between the
fingerprint samples [46], [47]. The general stages of a
fingerprint are shown in Figure 4.
During the fingerprint registration process, a sensor
scans the user’s fingerprint and converts it into a digital
image. Subsequently, the minutiae extractor analyzes
the fingerprint image to identify distinct details known
as minutiae points. In the user verification stage, the
same sensor is touched again, and a new fingerprint
image is created. This new image is called the query
Print. The system extracts and compares the minutiae
points from this image with the minutiae points stored
in the database to find the number of common minutiae
points. Due to differences in pressure, the query image
and the database image need to be registered before
comparison. Registration refers to aligning the two
images to match the corresponding minutiae pairs.
Then, the matching unit calculates the number of
matching pairs between the two samples. Matching
pairs refer to the minutiae points that have the same
location and direction.
Upon completing the previous stage, the system
determines the user’s identity by calculating the match-
ing score and comparing it to a predefined thresh-
old value. This process allows for the determination
of whether the user’s fingerprint sufficiently matches
the reference template for successful authentication.
In this method, the direction of the ridges and their
frequency are first determined, which improves the
quality of the image and facilitates ridge extraction.
After image enhancement, the main skeleton of the
friction ridges is extracted, and finally an algorithm
is used to identify and eliminate false minutiae using
heuristics. Then, the matching phase begins. In the
matching stage, as explained earlier, a score is calcu-
lated for the comparison of two fingerprints. One of
the most challenging parts of fingerprint authentication
is this matching stage because there are intra-class
differences that indicate variations between different
images of the same fingerprint, and there are inter-class
similarities within the fingerprint. For example, factors
such as finger pressure, finger placement, and rotation
affect intra-class differences. Also, the existence of only
9
The Role of Biometric in Banking: A Review
EAI Endorsed Transactions on
AI and Robotics

Figure 4. Stages of Identity Verification through
Fingerprint
three general fingerprint pattern types (ring, spiral, and
arch) creates similarities between different groups. In
the following, we will examine the research presented
in the field of identification methods, and at the end,
we will compare them.
Yang, et al. [48] have conducted research on
fingerprint matching based on maximum training. To
match the fingerprint, this system uses ELM (Extreme
Learning Machine) and R-ELM (Regularized Extreme
Learning Machine). This approach is designed to
address the shortcomings of conventional learning
methods. The method involves several key steps,
including efficient preprocessing, extraction of moment
features, and principal component analysis for feature
selection. The research findings demonstrate that this
proposed method achieves superior matching accuracy,
requiring less time compared to traditional approaches.
The ELM and R-ELM methods introduced in this
study prove to be more effective than the conventional
methods.
Labati, et al. [49] have presented research on the
extraction of minutiae in heterogeneous fingerprint
images using artificial neural networks. In this method,
sweat pores are utilized for quality assessment, and
a technique for extracting minutiae coordinates from
touch, touchless, and unstable fingerprint images is
proposed. Specifically, neural networks are designed
and trained to generalize the centrality of each
minutia. The results of this study demonstrate that
this approach achieves higher accuracy and better
performance compared to other methods (such as
extraction methods based on minutiae without neural
network-based information classification).
FengJ et al. [50] have presented research on
highly complex neural networks for enhancing latent
fingerprints. This method utilizes two components,
namely the ridge and non-ridge (smooth) regions. The
ridge region is used for enhancing fingerprint feature
extraction, while the non-ridge region is employed for
orientation estimation and noise removal. The neural
network is trained using a multitask and pixel-to-
pixel approach. The results of this study demonstrate
that this method achieves desirable outcomes for
incompatible and latent fingerprints.
Khodadoust, et al. [51] have delved into a research
endeavor centered around fingerprint representation,
employing an innovative approach termed triangular
expansion. In this pursuit, the method zeroes in on
the utilization of triangular expansion for fingerprint
representation, imparting a distinct and effective
dimension to the study. By proposing an algorithm
that leverages the power of a triple feature minutiae
representation, the authors have set out to elevate
the performance of fingerprint indexing. These three-
dimensional feature vectors, which draw inspiration
from molecules, serve as the foundation for generating
indices that play a pivotal role in the intricate
matching process. The heart of their approach rests
upon the incorporation of an enhanced K-MEANS
algorithm, a sophisticated mechanism that refines the
process by which fingerprint candidates are selected
for comparison. This strategic refinement paves the
way for a significant reduction in the initial candidate
list, ultimately culminating in the construction of a
well-curated final candidate list that holds paramount
importance during the matching phase.
On a parallel track, Zhao C. et al. [52] have
contributed substantially to the realm of fingerprint
verification, steering their research compass toward
the integration of fuzzy noise fingerprints within the
physical layer of wireless networks. This deliberate
exploration is anchored in the notion of enhancing
not only the security but also the overall efficiency of
wireless network operations. By skillfully leveraging
the concept of fuzzy noise fingerprints, the authors
have harnessed the potential to fortify the security
mechanisms of wireless networks, underlining their
commitment to a more robust and resilient digital
landscape.
10
Marani et al.
EAI Endorsed Transactions on
AI and Robotics

Wireless communication systems face an ongoing
challenge of unwanted data transmission from invalid
sources, often due to unauthorized actors. In response,
the authors present an innovative solution—an authen-
tication algorithm using fuzzy noise fingerprints. This
algorithm not only verifies identities but also ensures
data integrity. By incorporating fuzzy noise finger-
prints, it safeguards identity verification processes and
defends against network attacks. The inclusion of com-
bined PHY fingerprints enhances network stability, for-
tifying wireless communication systems against disrup-
tions.In their independent research, Wang et al. [53]
introduce a novel and inventive strategy for crafting
non-cancelable fingerprint templates. The main objec-
tive of this approach is to develop templates that
provide robust protection to the original fingerprint
data, preventing any possibility of deletion or loss and
ensuring that a new pattern or template for the finger-
print cannot be accepted.Unlike cancelable templates,
which may require initial image alignment and can
be susceptible to incorrect detection of specific points,
non-cancelable templates offer distinct advantages. The
proposed technique introduces a localized construction
process where Minutiae structures are formed through
semi-zone pairs, enabling the creation of removable
fingerprints without the need for template alignment.
Additionally, a discrete relative feature transformation
based on the Fourier transform is proposed. These new
templates fulfill the requirements of diversity, cancella-
bility, and irreversibility. The performance evaluation
of these templates includes testing on multiple pub-
lic databases, demonstrating exceptional performance
with the same error rate for lost fingerprints.
The research conducted by Zhao C. et al. and
Wang S et al. makes a significant contribution to the
advancement of fingerprint verification and template
design. Their studies present innovative solutions to
tackle vital challenges in wireless network security and
safeguarding fingerprint data. The proposed algorithms
and techniques provide promising results, highlighting
the potential for improving the reliability and resilience
of fingerprint-based authentication systems in various
applications. A detailed review and comparison of
fingerprint recognition methods are presented in Table
3.
3.4. Processor
In the field of embedded systems and microcontroller
programming, there exist various types of protocols
that facilitate communication and interaction between
different components. Some of the commonly encoun-
tered protocols include ARM, AVR, PIC, and FPGA. A
detailed review and comparison of these protocols are
presented in Table 4.
AVR: AVR, short for Alf and Vegard’s RISC proces-
sor, constitutes a family of microcontrollers created by
Atmel (now owned by Microchip Technology). These
microcontrollers are built on the principles of Reduced
Instruction Set Computing (RISC) architecture and find
extensive usage across diverse applications, including
consumer electronics, industrial automation, and Inter-
net of Things (IoT) devices. AVR microcontrollers are
known for their low power consumption, high perfor-
mance, and ease of use. They offer a range of fea-
tures such as analog-to-digital converters, timers, serial
communication interfaces, and programmable I/O pins.
These microcontrollers offer a compelling combination
of low power consumption, high performance, ease of
use, and a rich set of features. These characteristics
make them a preferred choice for a wide range of
applications, where power efficiency, reliability, and
flexibility are paramount. The AVR family continues to
be a significant player in the microcontroller market,
driving innovation and powering numerous electronic
systems around the world.
ARM: ARM (Advanced RISC Machines) is a fam-
ily of microprocessor architectures developed by ARM
Holdings (now owned by NVIDIA). ARM processors
are extensively utilized in mobile devices, embedded
systems, and various other applications. Renowned
for their energy efficiency, scalability, and adaptability,
ARM processors have found their way into a diverse
array of devices, such as smartphones, tablets, smart-
watches, and automotive systems. They offer different
cores and instruction sets to meet the requirements
of different applications, ranging From energy-efficient
IoT devices to high-powered computing systems. These
kinds of processors offer a compelling combination
of energy efficiency, scalability, and flexibility. Their
widespread adoption in mobile devices, embedded sys-
tems, and diverse applications is a testament to their
effectiveness. The ARM architecture continues to drive
advancements in the industry, enabling efficient and
powerful computing solutions for an increasingly con-
nected world.
PIC: Microchip Technology’s PIC (Peripheral Inter-
face Controller) is a family of microcontrollers, known
for their versatility and widespread use in embedded
systems. These microcontrollers offer a diverse range of
features, including timers, analog-to-digital converters,
serial communication interfaces, and digital I/O pins.
Renowned for their user-friendly nature, affordability,
and the abundance of development tools and libraries,
PIC microcontrollers are extensively applied in various
domains such as industrial control, home automation,
medical devices [54] , and consumer electronics.
FPGA: FPGA (Field-Programmable Gate Array) is a
type of programmable integrated circuit that stands out
for its flexibility, as users can configure and reconfigure
it as needed after the manufacturing phase. FPGAs
11
The Role of Biometric in Banking: A Review
EAI Endorsed Transactions on
AI and Robotics

Table 3. Comparison of Fingerprint Recognition Methods
Method Advantages Disadvantages
Ordinary •Ease of execution
•Extensive data classification is minimally
required
•The process execution is lengthy
•The accuracy during the alignment stage is low
Neural Network •Increased accuracy in all stages
•Increased efficiency in fingerprint registration
and verification processes.
•Low fingerprint quality captured by the sensor
can lead to detection issues.
• High memory requirement for registration in
different stages
• Dependency on capturing all points of the
fingerprint during scanning
Hybrid •It maintains proper functionality in case of
incomplete fingerprints.
•High speed in recognition due to utilizing only a
portion of the scanned image.
•No need for extensive memory.
•Greater complexity compared to other tech-
niques.
Table 4. Comparison of Microcontroller Types
P rocessor Su bset siz e f requency Learning S our ces P ri ce Gener al P o wer S pecial ized power N ois e P rot ocol
AVR Over 120 300 MHz very high affordable moderate low high Moderate
ARM Over 200 over 1 GHz moderate moderate high high low Very good
PIC Over 60 40 MHz high moderate moderate moderate low Good
FPGA Over 200 over 1 GHz moderate moderate moderate high low Moderate
consist of programmable logic blocks and interconnects
that can be configured to implement digital logic
circuits. Unlike microcontrollers or microprocessors,
FPGAs offer a high level of flexibility and can be
customized to perform specific tasks or algorithms.
They are used in applications that require high
performance, parallel processing, and real-time data
processing, such as digital signal processing, image
and video processing, graph processing [55],[56], and
scientific research. All these are different technologies
used in the field of microcontrollers and digital logic
design. Each has its own strengths and applications,
some of which have been presented in Table 4.
4. Conclusion
In conclusion, this review paper has provided a
comprehensive examination of the current state of
fingerprint recognition devices in ATMs, shedding
light on key research contributions in the field
of biometric authentication. Through a thorough
evaluation of fingerprint recognition algorithms, we
have identified areas where improvements can be made,
specificallywith regards to reliability, robustness, and
proficiency.
Additionally, we have tackled the weaknesses con-
nected with biometric authentication in ATMs, empha-
sizing the significance of bolstering security protocols.
An avenue for potential improvement lies in employ-
ing transaction graphs and graph-based algorithms.
Through constructing transaction graphs and apply-
ing such algorithms, researchers can efficiently identify
suspicious activities, such as fraudulent transactions or
unauthorized access attempts. These valuable findings
can be harnessed to elevate the security measures inte-
grated into ATM authentication systems [57], [58].
By bridging the gap between theoretical advance-
ments and practical implementation, this review paper
contributes to the broader understanding of fingerprint
12
Marani et al.
EAI Endorsed Transactions on
AI and Robotics

recognition devices in ATMs. It provides a founda-
tion for future researchers and industry professionals
to build upon, working towards the common goal of
enhancing the security and efficiency of ATM authen-
tication systems [59], [60].
As the field continues to evolve, it is expected
that further advancements will be made, both in
terms of biometric technology and security measures.
This review paper serves as a catalyst for future
exploration and innovation in the realm of biometric
authentication in the banking industry. By leveraging
the insights gained from this review, researchers and
industry experts can collaborate to develop more secure
and reliable ATM authentication systems, ultimately
safeguarding the interests of customers and financial
institutions alike.
References
[1] T. Sabhanayagam, V. P. Venkatesan, and K. Senthama-
raikannan, "A comprehensive survey on various biometric
systems," International Journal of Applied Engineering
Research, vol. 13, no. 5, pp. 2276-2297, 2018.
[2] S. Khan, S. Parkinson, L. Grant, N. Liu, and S. Mcguire,
"Biometric systems utilizing health data from wearable
devices: applications and future challenges in computer
security," ACM Computing Surveys (CSUR), vol. 53, no. 4,
pp. 1-29, 2020.
[3] M. Gayathri, C. Malathy, and M. Prabhakaran, "A review
on various biometric techniques, its features, methods,
security issues and application areas," Computational
Vision and Bio-Inspired Computing: ICCVBIC 2019, pp.
931-941, 2020.
[4] I. M. Alsaadi, "Study on most popular behavioral biomet-
rics, advantages, disadvantages and recent applications: A
review," Int. J. Sci. Technol. Res, vol. 10, pp. 15-21, 2021.
[5] W. Yang, S. Wang, N. M. Sahri, N. M. Karie, M. Ahmed,
and C. Valli, "Biometrics for internet-of-things security: A
review," Sensors, vol. 21, no. 18, p. 6163, 2021.
[6] A. F. Y. Althabhawee and B. K. O. C. Alwawi, "Fingerprint
recognition based on collected images using deep learning
technology," IAES International Journal of Artificial
Intelligence, vol. 11, no. 1, p. 81, 2022.
[7] Ansaripour N, Heidari A, Eftekhari SA. Multi-objective
optimization of residual stresses and distortion in
submerged arc welding process using Genetic Algorithm
and Harmony Search. Proceedings of the Institution of
Mechanical Engineers, Part C: Journal of Mechanical
Engineering Science. 2020 Feb;234(4):862-71.
[8] Daniali OA, Toghraie D, Eftekhari SA. Thermo-hydraulic
and economic optimization of Iranol refinery oil heat
exchanger with Copper oxide nanoparticles using
MOMBO. Physica A: Statistical Mechanics and its
Applications. 2020 Feb 15;540:123010.
[9] Babajamali Z, Aghadavoudi F, Farhatnia F, Eftekhari
SA, Toghraie D. Pareto multi-objective optimization of
tandem cold rolling settings for reductions and inter
stand tensions using NSGA-II. ISA transactions. 2022 Nov
1;130:399-408.
[10] Barnoon P, Toghraie D, Mehmandoust B, Fazilati MA,
Eftekhari SA. Natural-forced cooling and Monte-Carlo
multi-objective optimization of mechanical and thermal
characteristics of a bipolar plate for use in a proton
exchange membrane fuel cell. Energy Reports. 2022 Nov
1;8:2747-61.
[11] Khan MA, Arshad H, Khan WZ, Alhaisoni M, Tariq
U, Hussein HS, Alshazly H, Osman L, Elashry A. HGR-
BOL2: human gait recognition for biometric applica-
tion using Bayesian optimization and extreme learning
machine. Future Generation Computer Systems. 2023 Jun
1;143:337-48.
[12] Soltani M. Joint optimization of opportunistic predictive
maintenance and multi-location spare part inventories
for a deteriorating system considering imperfect actions.
arXiv preprint arXiv:1810.06315. 2018 Oct 15.
[13] Soleimani M, Forouzanfar Z, Soltani M, Harandi
MJ. Imbalanced Multiclass Medical Data Classification
based on Learning Automata and Neural Network. EAI
Endorsed Transactions on AI and Robotics. 2023 Jul 24;2.
[14] Sun H, Zhang Y, Chen P, Wang H, Liu YP, Liang R.
A New Approach in Automated Fingerprint Presentation
Attack Detection using Optical Coherence Tomography.
IEEE Transactions on Information Forensics and Security.
2023 Jul 7.
[15] M. N. Kumar, S. Raghul, K. N. Prasad, and P. N. Kumar,
"Biometrically Secured ATM Vigilance System," in 2021
7th International Conference on Advanced Computing
and Communication Systems (ICACCS), 2021, vol. 1:
IEEE, pp. 919-922.
[16] K. M. Htwe, Z. M. M. Htun, and H. M. Tun,
"Design and implementation of bank locker security
system based on fingerprint sensing circuit and RFID
reader," International Journal of Scientific and Technology
Research, vol. 4, no. 7, pp. 6-10, 2015.
[17] K. R. Gill and J. Sachin, "Vehicle ignition using
fingerprint sensor," Int. J. Innov. Res. Sci. Technol, vol. 2,
no. 12, pp. 357-363, 2016.
[18] R. Madhu and U. Mahadevaswamy, "GSM/GPS based
Device Switching with Fingerprint Module Integration
using Arduino," International Journal of Computer
Applications, vol. 149, no. 2, 2016.
[19] A. J. Moshayedi, M. Hosseinzadeh, B. P. Joshi, M.
Emadi Andani,(2023). Recognition System for Ergonomic
Mattress and Pillow: Design and Fabrication. IETE Journal
of Research, 1-19.
[20] S. Singh, A. Singh, and R. Kumar, "A constraint-based
biometric scheme on ATM and swiping machine," in 2016
international conference on computational techniques in
information and communication technologies (icctict),
2016: IEEE, pp. 74-79.
[21] S. I. Manzoor and A. Selwal, "An analysis of biometric
based security systems," in 2018 Fifth International
Conference on Parallel, Distributed and Grid Computing
(PDGC), 2018: IEEE, pp. 306-311.
[22] K. M. Sagayam, D. N. Ponraj, J. Winston, J. Yaspy, D. E.
Jeba, and A. Clara, "Authentication of biometric system
using fingerprint recognition with euclidean distance and
neural network classifier," Int. J. Innov. Technol. Explor.
Eng, vol. 8, no. 4, pp. 766-771, 2019.
13
The Role of Biometric in Banking: A Review
EAI Endorsed Transactions on
AI and Robotics

[23] Z. Ishak, N. Rajendran, O. I. Al-Sanjary, and N. A.
M. Razali, "Secure biometric lock system for files and
applications: a review," in 2020 16th IEEE International
Colloquium on Signal Processing, Its Applications
(CSPA), 2020: IEEE, pp. 23-28.
[24] H. Hoorfar, A. Bagheri, Minimum hidden guarding of
histogram polygons. arXiv preprint arXiv:1708.05815.
2017 Aug 19.
[25] H. Hoorfar, Bagheri A. A New Optimal Algorithm for
Computing the Visibility Area of a simple Polygon from
a Viewpoint. arXiv preprint arXiv:1803.10184. 2018 Mar
27.
[26] M. Ferretti, S. Nicolazzo, and A. Nocera, "H2O: secure
interactions in IoT via behavioral fingerprinting," Future
Internet, vol. 13, no. 5, p. 117, 2021.
[27] X. Yin, S. Wang, M. Shahzad, and J. Hu, "An IoT-oriented
privacy-preserving fingerprint authentication system,"
IEEE Internet of Things Journal, vol. 9, no. 14, pp. 11760-
11771, 2021.
[28] T. Nonthaputha, M. Kumngern, S. Kaewwang, J.
Phookwantong, N. Thepnarin, and K. Anontree, "A Design
of Smart Biometric Fingerprint Key Storage Cabinet Based
on IoT," in 2022 20th International Conference on ICT and
Knowledge Engineering, 2022: IEEE, pp. 1-5.
[29] V. M. M. Siregar and N. F. Siagian, "The Implementation
of Fingerprint Sensors for Fingerprint Reader Prototypes
Using a Microcontroller," Internet of Things and Artificial
Intelligence Journal, vol. 2, no. 1, pp. 47-59, 2022.
[30] M. Lastra, J. Carabaño, P. D. Gutiérrez, J. M. Benítez, and
F. Herrera, "Fast fingerprint identification using GPUs,"
Information Sciences, vol. 301, pp. 195-214, 2015.
[31] P. M. A. Hambalık, "Fingerprint recognition system
using artificial neural network as feature extractor: design
and performance evaluation," Tatra Mt. Math. Publ, vol.
67, pp. 117-134, 2016.
[32] A. J. Moshayedi, A. J. Li, N. Sina, Chen, X., Liao, L.,
Gheisari, M., Xie, X. (2022). Simulation and validation of
optimized pid controller in agv (automated guided vehi-
cles) model using pso and bas algorithms. Computational
Intelligence and Neuroscience, 2022.
[33] R. Murugesh, "Advanced biometric ATM machine
with AES 256 and steganography implementation," in
2012 Fourth International Conference on Advanced
Computing (ICoAC), 2012: IEEE, pp. 1-4.
[34] A. Taralekar, G. Chouhan, R. Tangade, and N. Shardoor,
"One touch multi-banking transaction ATM system using
biometric and GSM authentication," in 2017 International
Conference on Big Data, IoT and Data Science (BID), 2017:
IEEE, pp. 60-64.
[35] B. Saranraj, N. S. P. Dharshini, R. Suvetha, and K. U.
Bharathi, "ATM security system using Arduino," in 2020
6th International Conference on Advanced Computing
and Communication Systems (ICACCS), 2020: IEEE, pp.
940-944.
[36] H. A. Alzamel, M. Alshabanah, and M. Alsmadi, "Point of
Sale (POS) Network with Embedded Fingerprint Biomet-
ric Authentication," Hussah Adnan Alzame, Muneerah
Alshabanah, Mutasem K. Alsmadi," Point of Sale (POS)
Network with Embedded Fingerprint Biometric Authen-
tication", International Journal of Scientific Research in
Science and Technology (IJSRST), Online ISSN, 2019.
[37] A. V. Bataev, "The Model of Assessing Economic
Efficiency of Biometric ATMs," in 2019 IEEE Conference
of Russian Young Researchers in Electrical and Electronic
Engineering (EIConRus), 2019: IEEE, pp. 1365-1370.
[38] T. Sangeetha, M. Kumaraguru, S. Akshay, and M. Kan-
ishka, "Biometric based fingerprint verification system for
atm machines," in Journal of Physics: Conference Series,
2021, vol. 1916, no. 1: IOP Publishing, p. 012033.
[39] Moshayedi AJ, Roy AS, Liao L, Lan H, Gheisari M,
Abbasi A, Bamakan SM. Automation Attendance Systems
Approaches: A Practical Review. BOHR International
Journal of Internet of things, Artificial Intelligence and
Machine Learning. 2022 May 11;1(1):25-34.
[40] M. N. Kumar, S. Raghul, K. N. Prasad, and P. N. Kumar,
"Biometrically Secured ATM Vigilance System," in 2021
7th International Conference on Advanced Computing
and Communication Systems (ICACCS), 2021, vol. 1:
IEEE, pp. 919-922.
[41] C. Hochwarter, D. Jahnel, and A. Uhl, "Public Per-
ceptions, Preferences and Legal Aspects towards ATMs
with Biometric Authentication in Austria," in 2019 Inter-
national Conference of the Biometrics Special Interest
Group (BIOSIG), 2019: IEEE, pp. 1-6.
[42] A. J. Moshayedi, M. Soleimani, M. Marani, Sh.Yang,
A.Razi, M. Emadi Andani, (2023) Fingerprint Identifica-
tion Banking(FIB); Affordable and Secure Biometric IOT
Design. Paper presented at the International Seminar on
Artificial Intelligence, Networking and Information Tech-
nology (AINIT 2023), China.
[43] K. O. Okokpujie, F. Olajide, S. John, and C. G.
Kennedy, "Implementation of the enhanced fingerprint
authentication in the ATM system using ATmega128 with
GSM feedback mechanism," 2016.
[44] P. Subpratatsavee and N. Pubpruankun, "A Design and
Implementation of Attendance System Using Smallest
Wireless Fingerprint with Arduino Yún Embedded
Board," in Applied Mechanics and Materials, 2015, vol.
752: Trans Tech Publ, pp. 1057-1061.
[45] D. Maltoni, D. Maio, A. K. Jain, and S. Prabhakar,
Handbook of fingerprint recognition. Springer, 2009.
[46] Soleimani M, Naderi MH, Ashrafi AR. Tensor product of
the power graph of some finite rings. Facta Universitatis,
Series: Mathematics and Informatics. 2019 Mar 13:101-
22.
[47] Mahmudi F, Soleimani M. Some results on Maximal
Graph of a Commutative Ring, 2016.
[48] J. Yang, S. Xie, S. Yoon, D. Park, Z. Fang, and S.
Yang, "Fingerprint matching based on extreme learning
machine," Neural Computing and Applications, vol. 22,
pp. 435-445, 2013.
[49] R. D. Labati, A. Genovese, E. Munoz, V. Piuri, and F.
Scotti, "A novel pore extraction method for heterogeneous
fingerprint images using convolutional neural networks,"
Pattern Recognition Letters, vol. 113, pp. 58-66, 2018.
[50] J. Li, J. Feng, and C.-C. J. Kuo, "Deep convolutional
neural network for latent fingerprint enhancement,"
Signal Processing: Image Communication, vol. 60, pp. 52-
63, 2018.
[51] J. Khodadoust and A. M. Khodadoust, "Fingerprint
indexing based on expanded Delaunay triangulation,"
Expert Systems with Applications, vol. 81, pp. 251-267,
14
Marani et al.
EAI Endorsed Transactions on
AI and Robotics

2017.
[52] C. Zhao, M. Huang, L. Huang, X. Du, and M. Guizani,
"A robust authentication scheme based on physical-layer
phase noise fingerprint for emerging wireless networks,"
Computer networks, vol. 128, pp. 164-171, 2017.
[53] S. Wang, W. Yang, and J. Hu, "Design of alignment-
free cancelable fingerprint templates with zoned minutia
pairs," Pattern Recognition, vol. 66, pp. 295-301, 2017.
[54] A. J. .Moshayedi, D. C. Gharpure, (2017, May).
Evaluation of bio inspired Mokhtar: Odor localization
system. In 2017 18th international carpathian control
conference (ICCC) (pp. 527-532). IEEE.
[55] M. Soleimani, F. Mahmudi, M. H. Naderi, On the
Maximal Graph of a Commutative Ring. Mathematics
Interdisciplinary Research. 2021 Jul 2.
[56] Soleimani M, Mahmudi F, Naderi MH. Some results on
the maximal graph of commutative rings. Advanced
Studies: Euro-Tbilisi Mathematical Journal. 2023
Mar;16(supp1):21-6.
[57] Moshayedi AJ, Roy AS, Liao L, Lan H, Gheisari M,
Abbasi A, Bamakan SM. Automation Attendance Systems
Approaches: A Practical Review. BOHR International
Journal of Internet of things, Artificial Intelligence and
Machine Learning. 2022 May 11;1(1):25-34.
[58] Boroujeni SP, Pashaei E. A Hybrid Chimp Optimiza-
tion Algorithm and Generalized Normal Distribution
Algorithm with Opposition-Based Learning Strategy
for Solving Data Clustering Problems. arXiv preprint
arXiv:2302.08623. 2023 Feb 16.
[59] Mehrabi N, Boroujeni SP. Age estimation based on
facial images using hybrid features and particle swarm
optimization. In2021 11th International Conference on
Computer Engineering and Knowledge (ICCKE) 2021 Oct
28 (pp. 412-418). IEEE.
[60] Boroujeni SP, Pashaei E. Data clustering using chimp
optimization algorithm. In2021 11th international confer-
ence on computer engineering and knowledge (ICCKE)
2021 Oct 28 (pp. 296-301). IEEE.
15
The Role of Biometric in Banking: A Review
EAI Endorsed Transactions on
AI and Robotics