Fingerprint and Face Authentication Portable Digital Electronic Voting Machine

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Fingerprint and Face Authentication Portable Digital

Electronic Voting Machine


Brandon Sage Cagape John Carlo Lisondato Princess Diane Maboloc John A. Bacus*
College of Engineering College of Engineering College of Engineering College of Engineering
Education Education Education Education
Electronics Engineering Program Electronics Engineering Program Electronics Engineering Program Computer Engineering Program
University of Mindanao, Davao University of Mindanao, Davao University of Mindanao, Davao University of Mindanao, Davao
City, Philippines City, Philippines City, Philippines City, Philippines
b.cagape.513587@umindanao.ed johncarlolisondato1@gmail.com dianemaboloc@gmail.com john_bacus@umindanao.edu.ph
u.ph

Abstract—The paper presents the Electronic Voting Machine investing in Electronic Voting Machines (EVM). The other
(EVM) device to facilitate digital elections with the use of disadvantage of the Paper Ballot system is that it is susceptible
Raspberry Pi as its brain. The EVM is designed to authenticate the to fraud. There is an issue in the recent Philippine Presidential
voter through fingerprint or face authentication. The face Elections in which most voters were advised to voluntarily
authentication method is incorporated with an anti-spoofing surrender their ballots to the poll watchers as the machine that is
technique. Local Binary Pattern Histograms (LBPH) and Haar- supposed to read their ballots is under repair. Even if there are
Cascade Classifier models were used to detect and recognize faces no reported fraudulent activities, it is still possible; therefore,
and eye-blink counters for anti-spoofing. Both authentications
switching to fully automated voting is advisable to prevent fraud
were combined and incorporated into the voting graphic user
[2]. The Paper Ballot system can also hinder the voting ability
interface (GUI), programmed through Python and its libraries to
create the EVM, wherein the voter must authenticate through
of marginalized population, such as the illiterate, [pregnant]
fingerprint or face recognition if they failed the prior method. The women, Persons with Disabilities (PWDs), and senior citizens,
accuracy tests resulted in an overall 80% and 95.56% accuracy for since the voting interpretation and validity is left at the discretion
fingerprint and face authentication, respectively. The anti- of election officers. Electronic Voting Machines will prevent
spoofing technique resulted in an overall accuracy of 100%. The this, ensuring their votes are correctly counted [3]. Aside from
mock elections of 20 voters conducted to test the functionality of significantly reducing electoral costs, tackling fraudulent
the EVM resulted in an 80% fingerprint authentication success activities, and ensuring the participation of the marginalized
rate, while 100% of the remaining 20% passed the face population, using EVMs will also make vote counting quicker
authentication. Those who failed the fingerprint authentication since the results are declared within 2-3 hours compared to the
proceeded to face authentication, resulting in an overall 100% paper ballot system, which takes an average of 30-40 hours vote-
success rate in the authentication of all voters. The database counting time [4].
results matched the receipt of the voters, demonstrating the
device's overall efficiency of 100%. In conclusion, the paper A study of [5] uses PIC16F877A as a microcontroller in an
proposes an efficient and secure EVM that ensures fair and Electronic Voting Machine, making the voting process faster,
transparent digital elections. more reliable, and more efficient. However, there is a problem
with the fingerprint data since the database images have high
Keywords— Electronic Voting Machine, Local Binary Pattern resolution requiring more memory to be allocated. There is also
Histogram (LBPH), Haar-Cascade Classifier, Anti-Spoofing, an Arduino-based EVM that is particularly used in school
Raspbian OS, FirebaseDB elections. This EVM is workable but not 100% efficient since it
can only accommodate three (3) candidates and has limited
I. INTRODUCTION memory [6]. A face recognition feature is commonly used in an
Voting is one of the fundamental rights that an individual Electronic Voting Machine to authenticate the user registered on
possesses as an essence of democracy. It gives them the freedom the server. In the case of twin voters, a two-fold biometric
to choose candidates they think are most suited to represent the authentication is used to avoid it [4]. In biometric testing of the
people and their interests. Among the different voting systems study [7], the parameters False Acceptance Rate (FAR) and
in the world, the Philippines has adopted the Paper Ballot system False Reject Rate (FRR) are applied. The FAR is an
wherein the voters shade the circle beside the candidate’s name identification percentage used when there is an unexpected
on the ballots provided to them and are then fed into a machine acceptance of an unauthorized user, whereas the FRR is the
that will count their votes. This system is good to some extent; percentage of registered but rejected or unmatched users. Based
however, there are disadvantages. One is the possibility of ghost on the calculations, there is a FAR of 2%, an FRR of 10%, and
voting and the production and use of defective ballots, marked an overall 94% high accuracy indicating that the biometric is
as wastage. Despite the number of faulty ballots being a small reliable for user authentication. The OpenCV library is
percentage overall, this percentage is still significant, especially integrated and then executed by Python language. The Haar-
in small-scale voting [1]. This wastage can be avoided by Cascade and Local Binary Pattern Histogram model is used for
face detection and recognition authentication [8]. Also, the use middle, ring, pinky) they are to register with, they will also be
of Raspberry Pi 4 as the processing unit in running the face the only ones aside from the proponents who will know which
recognition was conducted [9]. A certain device with a face one to use for their authentication. This factor will add to the
recognition feature that also uses Haar-Cascade has a 92% security of the EVM system.
accuracy in recognizing its subjects. The Local Binary
This study proposes to create an Electronic Voting Machine
Histogram (LPB) algorithm resulted in an accuracy of 91%,
that uses two authentication methods, fingerprint and face
which both accuracy rates are considered high percentages [10].
recognition, to prevent fraud and the possibility of ghost voting.
One of the disadvantages of face recognition features is that it is
prone to identity fraud because of facial spoofing. Facial Producing an EVM will also remove the need to print out ballots
where there is unavoidable production of defective ones. The
spoofing is using a person’s face through a photo, a video, or a
EVM will be set in a kiosk style similar to closed Automatic
hyper-realistic mask to steal their identity [11]. This act can be
prevented by different anti-spoofing techniques such as using Teller Machine (ATM) places, making it easy to navigate for
first-time voters while giving them enough privacy when voting.
sensors or dedicated hardware for detecting facial features,
The only difference will be that the EVM kiosk will be lower
challenge-response methods of the user in real-time and
recognition algorithms (specular feature projections, depth and table-like compared to the ATMs to cater to PWDs,
especially those in a wheelchair. This adjustment will also make
feature fusion, image quality assessment, deep learning) [12].
the voters, especially the senior citizens, more comfortable when
The study focuses on an Electronic Voting Machine (EVM) voting. As a limitation, blind people are the only PWD excluded
with two authentication methods for authentic, transparent, and from being catered to because including them would yield
credible voting systems. Specifically, the EVM features additional features to be considered, which would cause
biometric authentication combined with face recognition additional expenses and complicated intricacies in the Electronic
authentication that will ensure the credibility of the voting Voting Machine. Compared to the usual voting system of the
system [13]. To completely authenticate the user, an anti- Philippines, using indelible ink after voting will no longer be
spoofing technique will be implemented together with face needed because the implementation is small-scale; the system
recognition to prevent identity theft through facial spoofing. The can check whether the individual has already voted.
technique that will be used is the challenge-response technique, Additionally, voter information and vote counts are saved on a
wherein the user is requested to interact with the system in a cloud database, presuming that there is a sudden and
specific way, such as doing facial expressions or head unprecedented data loss. A receipt will be printed after voting
movements [14]. An EVM is in the form of a portable for transparency and for the voters to ensure that their votes are
standalone tablet kiosk for convenience in voting. A receipt is read correctly. The receipts generated will then be dropped in a
printed after voting for the voters to ensure that their votes are drop box and kept by the committee for manual recounting,
read correctly. The receipt will be put in a drop box, which the especially when there are anomalies or a candidate requests it.
committee keeps for manual recounting in case a candidate files This aspect is another way of preserving data in case of
a petition for it [15]. corruption or loss. This study aims to implement small-scale
voting, such as school body elections and vote-based contests,
This study aims to develop a standalone Electronic Voting using this Electronic Voting Machine (EVM). Specifically, the
Machine (EVM) with two authentication systems. More implementation will occur at our university, The University of
specifically, this electronic voting machine should be able to Mindanao.
recognize and authenticate the fingerprints and faces, supported
by the anti-spoofing technique of the registered users as II. MATERIALS AND METHODS
implemented. This study also aims to perform an effective
voting procedure and accurately count the number of votes from A. Conceptual Framework
the registered users and tally them. Furthermore, the votes The study used Raspberry Pi as the main processing unit of
counted from the system should match the manual vote count the system, which ran the Raspbian operating system. The
from the receipts printed after the voting process. component used was compatible with the Adafruit fingerprint
This study would be of great importance as it will make scanner through its USB. Also, the Python language was used
voting more convenient with the aid of a device that will make for programming because of its compatibility with the libraries
the flow of the electoral process smoother. The Electronic used. The Open Computer Vision Library, or OpenCV, was
Voting Machine (EVM) will be a standalone computer kiosk to widely used for image processing and computer vision projects.
make the voting process ideal for the voters and will not be a Additionally, the library was utilized to build the Haar-Cascade
hassle. The monitor screen used for the voting process will have model for face detection/anti-spoofing and LBPH (Local Binary
bigger fonts to make the on-screen instructions very clear and Pattern Histogram) model for face recognition [8]. Aside from
this, the Pyqt5 library designed the system's graphical user
readable for people, especially senior citizens and those with
poor eyesight. With this, they can vote with little to no effort. interface (GUI).
The EVM will use fingerprint and face authentication features Fig. 1 shows the conceptual framework of the study.
to ensure the authenticity and credibility of the voting process. The system required fingerprint or face acquisition to validate
An anti-spoofing technique with a face recognition feature will the voter's identity for authentication, which was the system’s
be implemented to prevent identity theft through facial spoofing. input. The fingerprint acquired was preprocessed to optimize the
This technique will also prevent fraud as only the voter can sample and extract significant features to find the matched
access their registered account. Moreover, since the voter will fingerprint from the database [15-16]. If the system failed to
be the one to decide which fingerprint (left or right thumb, index, authenticate the registered voter's fingerprint, it proceeded to
face authentication. The system first performed face detection,
increasing the face recognizer's accuracy since it marked the
voter's face before proceeding to recognition. Afterward, the
face recognition algorithm matched the extracted voter's face
from the database. Moreover, the same face detection algorithm
was used for the anti-spoofing procedure. It detected the voter's
eyes and then required them to blink twice to prevent identity
theft. There were two authentication methods, but the voter was
only required to pass one of the biometric authentications before
proceeding to the voting proper, wherein the voter voted for their Fig. 3. LBPH (Local Binary Pattern Histogram)
preferred candidates among the selection and finalized their
votes. Furthermore, this comprised the electronic voting C. Voter’s Registration
machine. Before the election day, people eligible to vote were
registered ahead of time with their personal information, which
included their names and samples for fingerprint and face
recognition, the two methods of authentication for the EVM.
Furthermore, they were given informed consent regarding our
study. As shown in Fig. 4, voters must touch the fingerprint
scanner with their desired finger for authentication. The images
were processed with a highly sensitive pixel amplifier, then
extracted templates that were stored/saved in the database as a
unique ID.
Fig. 1. Conceptual Framework

B. Face Detection and Recognition Model


Fig. 2 shows the overall concept of the Haar-Cascade
classifier model. It used Haar features to determine the lines or
edges of the samples wherein there were changes in the pixel
intensity. It also used the integral images technique to summate
pixels over the image sub-area [10]. Aside from this, pre-trained
files were available to detect specific body parts like the face for
face detection. Eye detection requires the voter to blink their
eyes to prevent anti-spoofing.

Fig. 2. Ha-ar Cascade Classifier


Fig. 4. Fingerprint Registration Block Diagram
Fig. 3 illustrates the Local Binary Pattern Histogram or
LBPH model, one of the simplest algorithms used in facial In Fig. 5, 200 face images were collected from each voter
recognition. The model located the image structure and using a camera. The images were then pre-processed to provide
compared each pixel to its adjacent pixels. The adjacent values good-quality images rather than false ones. Next, feature
were then converted to binary based on a threshold, where 1s extraction was implemented to extract the face components.
represented values above the threshold, while 0s represented Lastly, images were trained by converting extracted features of
values below the threshold. The algorithm's accuracy was each image to binary, resulting in the center point being the
proportional to the number of samples in the dataset, and a decimal equivalent. These significant features were the face
minimum of 100 samples was required to achieve optimal encodings stored in pickle files used in face recognition.
accuracy [8]. In this study, 200 image samples were collected
from each voter.
When the voter fails to authenticate through fingerprint
recognition, it will proceed to face recognition. The voter's
images were scanned, and their facial features were extracted.
The system then compared the sample image with the stored
face encodings to determine the best match. If the sample image
matched the stored encodings, the voter’s name was selected,
and their information was fetched from the database. The system
then proceeded to the voting display.
If the voter fails again to authenticate through face
recognition, then the voter will not be able to vote.
E. System Integration
The EVM and registration were separated, as shown in Fig.
7. A laptop was used to collect voter data, including their names
and authentication data. The gathered data was sorted and
transferred to the storage of the Raspberry Pi. The transferred
features were then used to authenticate the voters. Additionally,
the Internet of Things (IoT) database was used to store the users
who had already voted, together with their votes, in the Firebase
database to prevent data loss in case the Raspberry Pi storage
gets corrupted. Additionally, this cloud database only connected
Fig. 5. Face Registration Block Diagram when the EVM validated the voter’s already-voted status or
appended the recent voter and their vote selection.
D. Voter’s Recognition
To be able to vote, one had to pass either fingerprint or face
authentication. The voter was to authenticate first through the
fingerprint authentication method. However, if the voter’s
fingerprint biometric is not recognized, they will be redirected
to face authentication instead. If the voter failed to authenticate
themselves from both methods, they could not proceed to the
voting proper.
Fig. 7. System Integration
Fig. 6 shows the fingerprint and face recognition block
diagram. This diagram started when the voter touched the F. Hardware Development
fingerprint scanner for authentication. If successfully
Fig. 8 illustrates the prototype design of the Electronic
authenticated, the scanned fingerprint is enhanced through the
Voting Machine. The size and dimensions of the prototype were
HSP amplifier and matched to the stored fingerprint data. Once
based on the needs and capacity of the potential categories of
a match was found, the voter information was fetched from the
voters. There was a presumption that one of the potential
database for authentication, and then the recognized voter could
categories of voters who would use the Electronic Voting
proceed to the voting display.
Machine was Persons with Disabilities (PWDs), with the
exclusion of blind voters [17]. The Electronic Voting Machine
was placed on a table so that it reached the height of 48" (~1.2m),
which was the ideal height of an EVM according to the Kiosk
ADA Accessibility ADA Compliance for it to be accessible to
the PWDs, especially the people on a wheelchair like the
disabled.

Fig. 8. Electronic Voting Machine (EVM) Prototype Design

Fig. 6. Fingerprint and Face Recognition Block Diagram


G. Statistical Analysis TABLE II. CONFUSION MATRIX FOR FACE AUTHENTICATION
ACCURACY TEST RESULTS
This study employed two authentication methods to verify
the legitimacy of the voting process by conducting trials and
calculating the overall accuracy for fingerprint and face
authentication using a confusion matrix. Furthermore, precision,
which measured the accuracy of positive predictions, recall,
which measured the ability of the model to identify positive
cases, and F1 score, which provided a balanced measure of both
metrics, were computed. Additionally, the False Reject Rate
(FRR), the rate of registered voters that were not authenticated,
was compared to the precision. In contrast, the False Acceptance
Rate (FAR), the rate of unregistered voters authenticated with
another person's identity, was compared to the recall. Database
vote selection and voting receipt selection were compared.
III. RESULTS AND DISCUSSIONS Of the total trials for the registered voters, forty-five (45)
A. Electronic Voting Machine were authenticated (True Positive), and none (0) were rejected.
On the other hand, of the total trials for the unregistered voters,
Fig. 10 shows the actual device used to conduct the mock three (3) were falsely identified as other people (False Negative),
election with the following components used: the monitor screen,
while forty-two (42) were classified as unknown (True
mouse, camera, and thermal printer, all of which were connected
Negative).
to the Raspberry Pi 4 to perform the voting procedure.
From the accuracy testing, the face recognition algorithm
had a confidence value of 97%. The resulting confidence value
of the registered voters ranged from 98.09% to 99.82%, while
the unregistered voters ranged from 89.98% to 97.98%. With
this, the algorithm confidence level was tuned to 98.09, adjusted
from the lowest confidence value of the registered voters to
prevent false negative results on the actual mock election.
The overall accuracy of the face authentication method is
95.56% using equation (1). The precision and recall are 100%
and 91.84% using equations (2) and (3), respectively. Lastly, the
F1-score resulting from equation (4) was 95.75%. Moreover, the
Fig. 9. Actual Device False Rejection Rate (FRR) resulted in 0% using equation (5),
while the False Acceptance Rate (FAR) resulted in 8.16% using
B. Accuracy Testing Results equation (6).
The accuracy test results for the fingerprint authentication
are shown in Table I. Fifteen (15) samples were tested, with
three trials conducted for each sample. The overall accuracy was
calculated by taking the percentage average of the three trials

TABLE I. FINGERPRINT AUTHENTICATION ACCURACY TEST RESULTS

Table III shows the overall accuracy of the anti-spoofing


technique (blink twice) used after a successful face
authentication. The overall accuracy was computed by the
success count over the total number of trials in which the result
was 100%.
Based on the results, a low accuracy of 60% was observed
in the first trial. During the testing, some fingers sampled in the TABLE III. ANTI-SPOOFING TEST RESULTS
first trial had wet or dirty fingers, which caused them not to be
recognized. After noticing the problem, they clean their fingers,
increasing accuracy in the next two trials.
The accuracy test results for face authentication are shown
in Table II, presented in a confusion matrix with two
classifications: registered and unregistered voters. Thirty (30)
samples were used, with three trials conducted for each sample.
C. Mock Election Results IV. CONCLUSIONS AND FUTURE WORKS
The results in Fig. 10 showed the type of authentication in A Raspberry Pi-based IoT portable electronic voting
which the voters have successfully verified. It is marked as machine system was successfully designed, incorporating
“Successful” if the voter has authenticated in that authentication fingerprint and face authentication methods for voter identity
method, while an “Error” is marked when the voter has failed to verification. The Adafruit fingerprint sensor was used for
do so. In the case of the face authentication method, the voter is fingerprint authentication, achieving an overall accuracy of 80%.
marked as “Passed” when it already has successfully The lower accuracy threshold was attributed to failed fingerprint
authenticated with their fingerprint, where they will no longer detection in the first trial, primarily due to wet or dirty
need to authenticate for their face. fingerprints. Face authentication, utilizing Haar-Cascade
Classifier and Local Binary Pattern Histogram (LBPH) models,
Based on the results, all voters were able to authenticate
achieved an overall accuracy of 95.56%. An anti-spoofing
themselves successfully. 16 out of 20 voters authenticated with
technique ensured 100% accuracy in detecting and preventing
their fingerprints, while 4 out of 20 voters had to go through face
spoofing attempts. While the face authentication method
recognition to authenticate themselves. The overall
successfully authenticated all registered voters, a few
accuracy is computed by the percentage of the fingerprint and
unregistered voters were mistakenly recognized, resulting in a
face authentication method's success rate. Given this, the
False Acceptance Rate (FAR) of 8.16%. In a mock election with
accuracy of fingerprint and 0face authentication are 80% and
20 registered voters, 80% were authenticated using the
20%, respectively, which resulted in an overall accuracy of
fingerprint authentication method, while the remaining 20%
100%.
proceeded to face authentication and achieved 100% overall
In addition, comparing the accuracy results of the fingerprint authentication accuracy. Vote results from the database and
authentication to the results of the prior accuracy testing, the voter's receipts matched with a 100% match for all 20
authentication method also had 80% accuracy. authenticated voters.
For future studies, the researchers recommend displaying the
user interface using a touchscreen LCD to reduce device
bulkiness and facilitate easier GUI navigation. Additionally, a
more sensitive fingerprint sensor with a high tolerance for wet
or dirty fingerprints is suggested.
REFERENCES
[1] ] K. M. AboSamra, A. A. AbdelHafez, G. M. R. Assassa, and M. F. M.
Mursi, “A practical, secure, and auditable e-voting system,” Journal of
Information Security and Applications, vol. 36. Elsevier BV, pp. 69–89,
Oct. 2017. doi: 10.1016/j.jisa.2017.08.002.
[2] R. M. Alvarez, I. Levin, and Y. Li, “Fraud, convenience, and e-voting:
how voting experience shapes opinions about voting technology,” Journal
of Information Technology & Politics, vol. 15, no. 2. Informa UK
Limited, pp. 94–105, Apr. 03, 2018. doi:
Fig. 10. Fingerprint and Face Authentication Results 10.1080/19331681.2018.1460288.
[3] R. Küsters and T. Truderung, “Security in e-voting,” it - Information
D. Database Votes and Receipt Votes Comparison Technology, vol. 56, no. 6. Walter de Gruyter GmbH, pp. 300–306, Nov.
30, 2014. doi: 10.1515/itit-2014-1062.
Based on the results in Fig. 11, all twenty (20) voters
[4] P. Katta, O. A. Mohammed, K. Prabaakaran, M. Divya, G. Jayashree, and
authenticated and proceeded to the voting proper successfully. D. Keerthika, “Smart voting using Fingerprint, Face and OTP Technology
The candidates chosen by each voter were stored in the EVM with Blockchain,” Journal of Physics: Conference Series, vol. 1916, no.
system, and the printed receipt containing the voter’s name and 1. IOP Publishing, p. 012139, May 01, 2021. doi: 10.1088/1742-
selected candidates were dropped in the drop box after 6596/1916/1/012139.
confirming their selections. The casted votes stored from the [5] A. I. Abeesh, P. Amal Prakash, R. Pillai Arun, H. S. Ashams, M. Dhanya,
database were compared to the receipts, and as presented in and R. Seena, "Electronic Voting Machine Authentication using
Biometric Information," in International Journal of Engineering Research
Table V, all votes matched from each other, resulting in an & Technology (IJERT), vol. 5, issue 16, pp. 1-6, Apr. 24, 2018. DOI:
accuracy of 100%. 10.17577/IJERTCONV5IS16012.
[6] Vinayachandra, K. G. Poornima, M. Rajeshwari, and K. K. Prasad,
TABLE IV. DATABASE VOTES AND RECEIPT VOTES COMPARISON “Arduino Based Authenticated Voting Machine (AVM) using RFID and
Fingerprint for the Student Elections,” Journal of Physics: Conference
Series, vol. 1712, no. 1. IOP Publishing, p. 012004, Dec. 01, 2020. doi:
10.1088/1742-6596/1712/1/012004.
[7] B. U. Umar, O. M. Olaniyi, L. A. Ajao, D. Maliki, and I. C. Okeke,
“Development of A Fingerprint Biometric Authentication System For
Secure Electronic Voting Machines,” Kinetik: Game Technology,
Information System, Computer Network, Computing, Electronics, and
Control. Universitas Muhammadiyah Malang, pp. 115–126, Mar. 04,
2019. doi: 10.22219/kinetik.v4i2.734.
[8] M. Bansal, "Face Recognition Implementation on Raspberry Pi Using
OpenCV and Python," International Journal of Computer Engineering &
Technology, vol. 10, no. 3, pp. 16-20, May 31, 2019. doi: & Security, vol. 26, no. 5. Elsevier BV, pp. 373–380, Aug. 2007. doi:
10.34218/ijcet.10.3.2019.016. 10.1016/j.cose.2006.12.011.
[9] E. K. M. Sanchez, N. B. Linsangan, and R. E. Angelia, “Three Triangle [15] Y. Lee, S. Park, M. Mambo, S. Kim, and D. Won, “Towards trustworthy
Method for Face Recognition using Dlib and OpenCV,” 2020 IEEE 12th e-voting using paper receipts,” Computer Standards & Interfaces,
International Conference on Humanoid, Nanotechnology, Information vol. 32, no. 5–6. Elsevier BV, pp. 305–311, Oct. 2010. doi:
Technology, Communication and Control, Environment, and 10.1016/j.csi.2010.03.001.
Management (HNICEM). IEEE, Dec. 03, 2020. doi: [16] [V. Malathy, N. Shilpa, M. Anand, and R. Elavarasi, “Radio frequency
10.1109/hnicem51456.2020.9400134. identification based electronic voting machine using fingerprint module,”
[10] S. M. Bah and F. Ming, “An improved face recognition algorithm and its IOP Conference Series: Materials Science and Engineering, vol. 981, no.
application in attendance management system,” Array, vol. 5. Elsevier 3. IOP Publishing, p. 032018, Dec. 01, 2020. doi: 10.1088/1757-
BV, p. 100014, Mar. 2020. doi: 10.1016/j.array.2019.100014. 899x/981/3/032018.
[11] S. Najam, A. Z. Shaikh, and S. Naqvi, “A Novel Hybrid Biometric [17] J. Lazar, J. B. Jordan, and G. Vanderheiden, “Toward unified guidelines
Electronic Voting System: Integrating Finger Print and Face for kiosk accessibility,” Interactions, vol. 26, no. 4. Association for
Recognition,” Mehran University Research Journal of Engineering and Computing Machinery (ACM), pp. 74–77, Jun. 26, 2019. doi:
Technology, vol. 37, no. 01. Mehran University of Engineering and 10.1145/3337779.
Technology, pp. 59–68, Jan. 01, 2018. doi: 10.22581/muet1982.1801.05.
[12] E. Fourati, W. Elloumi, and A. Chetouani, “Anti-spoofing in face
Your Name Title* Research Field Personal
recognition-based biometric authentication using Image Quality
Assessment,” Multimedia Tools and Applications, vol. 79, no. 1–2. website
Springer Science and Business Media LLC, pp. 865–889, Oct. 10, 2019. Brandon Cagape student Computer vision
doi: 10.1007/s11042-019-08115-w.
[13] S. L. Rikwith, D. Saiteja, and R. Jayaraman, “Enhancement of Electronic John Carlo student Computer vision
Voting Machine Performance Using Fingerprint and Face Recognition,” Lisondato
2021 2nd International Conference on Smart Electronics and
Communication (ICOSEC). IEEE, Oct. 07, 2021. doi: Princess student Computer vision
Maboloc
10.1109/icosec51865.2021.9591895.
[14] N. Komninos, D. D. Vergados, and C. Douligeris, “Authentication in a John Bacus Adviser Computer vision
layered security approach for mobile ad hoc networks,” Computers

You might also like