Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)

IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

Secured E-voting System Using Two-factor Biometric

Authentication
2020 Fourth International Conference on Computing Methodologies and Communication (ICCMC) 978-1-7281-4889-2/20/$31.00 ©2020 IEEE 10.1109/ICCMC48092.2020.ICCMC-00046

Sudeepthi Komatineni Gowtham Lingala

Dept of Computer Science & Engineering Dept of Computer Science & Engineering
Koneru Lakshmaiah Education Foundation Koneru Lakshmaiah Education Foundation
Vaddeswaram,India Vaddeswaram,India

Abstract --- Building a secure voting system that offers privacy detecting an individual’s face and then recognizing if it
of conventional voting system with proper voter authentication & matches with any of the voter’s face stored prior in the voter
transparency has been a challenge for a due course of time. The database thereby avoiding duplicate votes.
research work proposes a secured and robust electronic voting
system based on popular machine learning based facial
recognition algorithms and biometric authentication Face recognition undoubtedly stands to be the most
methodologies for the purpose of building a secure voting system.
demanding yet difficult computing method due to various
In particular, it focuses on the potential working of face detection
and recognition and bio-metric authentication namely bio-metric
factors like background of the image, image illumination, facial
scan, and the implementation procedure, which improves the expressions, angles and postures. Many algorithms are
security and decreases the duplicate vote and fraudulent to make available that tend to meet the above purpose yet Eigen-face
the system as more efficient and user friendly in nature. based facial recognition is considered to be more viable

Keywords— authentication; face detection and recognition; A. How it works?

biometric scanning; robust; user friendly The objective is to identify and differentiate the current
image of the individual with those that exist in the Voter
I. INT RODUCT ION database, a database where the biometric and the basic details
Democracy plays a vital role in many developing countries, of a registered voter are collected and stored prior. As
where the power lies in the hands of people and they decide mentioned above, the input image can be highly noisy at times
the government by the process of elections. Polling personnel or there might be a few changes in an individual’s latest facial
and the polling plays an indispensable role in elections. From features as the human face is a dynamic structure that can
the year of independence till today many changes happened in change rapidly with time, considering which the Eigen -face
the election procedures , initially ballet papers are used as approach stands out to be the efficient one, as in this approach
medium for votes and later it changed to electronic voting the comparison is made after anticipating an image based on
machines where the voters should choose from the list of the given image. In spite of the differences like noise or change
buttons where each represent a symbol. In due course of time in facial features there are unique characteristics named Eigen -
many measures are taken to reduce costs of election faces in every individual’s face like the presence of some
procedures without compromising on the voters’ privacy and objects (eyes, iris, nose, mouth, ears) and the relative distance
following procedures as instructed by constitution, but there between them. Eigen-faces can be extracted from any image
are still many cases registered pointing out the lack of proper through a mathematical approach called principal component
authentication procedures of voters. This work suggests analysis (PCA) by using which one can transform any image
methods that help in maintaining authentication and privacy of into its respective Eigen-faces, which are nothing but the
the voters by using bio-metric scanning [By utilizing bio- features of faces (like eyes, nose, mouth, ears). Though all of
metrics a voter could be distinguished based on their identity] them may not be present in the original/saved image at times,
face recognition methods using machine learning algorithms. adding up all the eigen-faces in the right proportion one can
reconstruct the original image as well, as eigen-faces are
nothing but the individual features of a person’s face. Features
II. FACE DETECTION AND RECOGNITION with bigger size in the original image will occupy a larger share
Face Recognition is leveraging as the most widely used in the total sum of Eigen-faces and features with smaller size or
biometric authentication for computing methods and Voter those doesn’t exist at all will occupy a smaller or no share in
fraud can be one of the most fruitful methods as it prevents the sum of Eigen-faces, this weight of eigen-faces specify to
individuals attempts of casting multiple votes under different what degree a particular feature is present in the original image,
names and ID’s. A major part of this issue can be resolved by thus by building weighted sum of all Eigenfaces the original
face can be reconstructed. This reconstructed image will be an
approximate original image in most of the cases because one
can obtain an exact original image only by reconstructing all

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 245

 Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)
IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

the extracted eigen-faces which is mostly not possible and C. PCA which plays a crucial role in this algorithm is a
only a part of the eigen-faces (important features) are generally method that finds an arrangement of projection vectors
used to limit the computation and computational resources (eigen vectors) comparing to the Eigen estimations of the
thereby resulting in an approximation of the original. original image, performed as follows:
1) Prepare the Voter Dataset:
x All the images (must be of same size) of the registered
voters stored in dataset are to be converted into their
respective vectors and each vector will have its
respective eigen-faces.
x Say there M images I1, I2, I3, ............... IM each of size
NxN and are converted respectively into vectors Г1, Г2,
Г3, ГM.
2) Calculating the mean of Image-vectors:
x Calculating mean of the image-vectors is used in
further calculation of eigen-vectors (oreigen-faces)

3) Calculating the Covariance matrix:

x Initially the average of image-vectors is subtracted
from the image needed say Гi

x Now covariance matrix C is obtained by the following

formula:

Fig. 1. Working of Eigen Face Recognition Algorithm

4) Calculating Eigen-vectors or Eigen-Faces from the

covariance matrix:
x Among the multiple matrix obtained from above step
consider one matrix say L and
B. The additional features of Eigen-face based algorithm
L vi = μvi, where vi is an individual eigen-vector
include:
x Determining if the image is a face, if the weights of x Now the total Eigen-vectors of ui the image vector i
current image reconstructed vary too much from the i.e., the image ith Image are given by the formula:
weights of all face’s images, then there are high chances
that the current image might not be a face at all. Thus,
unnecessary comparison of this current image with the
images of datasets is avoided.
x Thus, the Eigen-faces can be calculated. Consider only
x All the similar images/faces of the database can be the larger eigen-face values and omit the lower ones as
grouped together, similar faces will have similar major characteristics of the face should only be
features thus similar eigen-faces can be formed with considered. In a similar way eigen-faces and also
almost equal weights resulting in an approximately weighted eigen-faces sum of every image of a
equal weighted-eigen sum. This while checking a registered voter in the database should be calculated and
current image, the eigen-faces can be extracted first stored. Images having approximately equal or similar
finding the weighted eigen-sum and then this current weighted-sum can be clustered together so that
image can be checked with images of dataset that computation can be made easy while checking. While
belong to cluster with the sum that is almost equal to the an individual is trying to cast his/her vote, eigen-faces
earlier. Thus, computation can be decreased as the and weighted-eigen sum of that image is calculated
whole dataset is not tested.

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 246

 Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)
IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

using the above process and then find out the weighted- d) Minutiae Matching: Minutiae matching is used to
Eigen sum of images in the database closer to the compare the data stored in the database with the test data using
current weighted eigen-sum and then compare the the below formulae :
eigen-faces of current image with those of the closest
cluster in the database. If a match is found then the
individual is a legal voter else there are high chances of
voter fraud.
x Though facial recognition can do mass identification
there might be cases where this algorithm may be Where xi represents minutiae points of references stored in
fooled like people with same facial features or identical database and Xi represents minutiae points of the voter’s
twins. Hence, it's not the most reliable and efficient
fingerprint that is to be tested or verified.
biometric authentication techniques at times when
compared to others. Thus, in order to make the Voting
system more secure and efficient a two-step
verification of 1. Facial Recognition and 2. Fingerprint
recognition is needed which can never be failed. Matching score varies on a range of 0 to 1 indicating a
complete fingerprint mismatch and a perfect fingerprint match
III. THUMB RECOGNITION respectively.
Fingerprint recognition is the most unique biometric
authentication known so far, as the ridge arrangement on every
finger of every individual is unique and it will not alter
although the dimensions of fingers change with respect to
growth and age. There is never a scenario where any two
fingerprints among billions available are found similar so far.
Thus a 2-factor verification using fingerprint recognition will
make the system more efficient and foolproof. Out of all the
algorithms available minutiae algorithm is widely used for
fingerprint recognition and is described as follows:

1) Binarizing the Fingerprint Image:

The usual fingerprints taken from an individual are called Fig 1. The above figure represents grey scale image, binary
the grey scale images with intensity ranging from 0 to 255, image, ridge thinned image with minutiae marked
which makes it vivid that they aren’t clear enough to mark respectively
minutiae. Hence, they are converted into binary images. A
threshold value is fixed and the pixel values that fall above it
are set to 1 and below it to 0 respectively which implies that the
Binarized Images are more intense.
2) Ridge Thinning: Though the Binarized image is intense
it isn’t thin enough to extract minutiae effectively, so it is
thinned to reduce the thickness of all ridges to single pixel
width using block filter. Ridge thinning doesn’t alter the
location of minutiae, it just places white pixels at the
boundaries of the ridges of a binary image preserving the
outermost pixels.

3) Minutiae Extraction & Marking: The ridged image is Fig 2. Working of Minutiae based Fingerprint Recognition
further divided into 3x3 matrices and crossing number Algorithm
technique described below is applied to mark the minutiae as:
a) Bifurcation: In a 3x3 matrix if the central pixel and 3 During the time of registration face images and fingerprints
of its neighbors have a value of 1 then the central pixel is a should be taken along with the basic voter details, such that at
branch of bifurcation. the time of polling a two-factor authentication of the voter can
b) Termination: If central pixel and just one of its be done by using Eigen-face based algorithm for facial
neighbors have a value of 1 then the central pixel is a ridge recognition and minutiae algorithm for fingerprint recog nition.
ending or termination. The voter will be allowed to cast his vote only if the eigen -
c) Trifurcation: If both the uppermost pixel and the faces and fingerprints of the voter match with those in the
rightmost pixel have a value of 1 false minutiae points are database. In case of physically challenged people, who
removed and left-over minutiae points are marked.

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 247

 Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)
IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

cannot submit their fingerprints, iris and facial recognition can

be done. This secure system thus avoids the vote frauds and CONCLUSION
malpractices during elections.
A 2-factor Bio-metric authentication system - 1. Face
recognition using Eigen face-based recognition algorithm
If this computing method can be developed as a highly secure and 2. Minutiae based algorithm with a full pledged
software as well then this can be used for both online voting database as an input transforms the traditional voting system
and traditional polling booths thereby avoiding malpractice & into a better and fraud free one. Also the above computing
frauds, minimizing time and efforts taken to cast the votes and method can be developed as a highly secure software as
counting them and by implementing online voting, polling well, then this can be used for both online voting and
percentage can be increased tremendously which had been a traditional polling booths thereby avoiding malpractice &
major throwback of elections in India over decades. frauds, minimizing time and efforts taken to cast the votes
and counting them and by implementing online voting,
polling percentage can be increased tremendously which
had been a major throwback of elections in India over
decades. Errors in election procedures can be a threat to
democracy, it marginalizes vast majority of voters causing
low percent of voter turning out and general political
ignorance, ignorance isn’t the problem but the problem is
the cause for ignorance.

REFERENCES

[1] Kaur, R., & Himanshi, E. (2015). Face recognition using Principal
Component Analysis. 2015 IEEE International Advance Computing
Conference (IACC). doi:10.1109/iadcc.2015.7154774 .
Fig 3. Working of Secure E-voti ng system using two-
[2] Zhao, X., & Wei, C. (2017). A real-time face recognition system based
factor authentication on the improved LBPH algorithm. 2017 IEEE 2nd International
Conference on Signal and Image Processing
I V. Advantages over the existing system (ICSIP). doi:10.1109/siprocess.2017.8124508
The conventional voting system have plethora of problems, [3] Mil’shtein, S., Pillai, A., Shendye, A., Liessner, C., & Baier, M.
(2008). Fingerprint Recognition Algorithms for Partial and Full
Procedure we follow is time taking, involve more man Fingerprints. 2008 IEEE Conference on Technologies for Homeland
power huge costs and yet not secure, many complaints get Security. doi:10.1109/ths.2008.4534494
registered for every voting cycle organized, vote is the basic [4] Finger Print Algorithm ( https://www.supremasolution.com/T ech2.php)
right of every citizen of the country, due to malpractices [5] Mallipamula Megha Sai Sree et al, International Journal of Computer
that are getting arise leads to the demise of rights of the Science and Mobile Computing, Vol.3 Issue.8, August - 2014, pg. 476-
citizens, major malpractices that occur during the election 481” Finger Print Enhancement Using Minutiae Based Algorithm”
[6] Manisha Redhu and Dr.Balkishan/ International Journal of Engineering
procedures are mainly due to failures of manual Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
authentication the solution proposed can solve the issue of Vol. 3, Issue 4, Jul-Aug 2013,pp .2488-2497 2488 | P a g e Fingerprint
authentication, by dual authentication method the voters Recognition Using Minutiae Extractar
bio-metrics and face are recognized by algorithms that are [7] Bindhu, V. "Biomedical Image Analysis using Semantic
robust, these help in solving the major cases of voter Segmentation." Journal of Innovative Image Processing (JIIP) 1, no. 02
(2019): 91-101.
authentication. During the time of registration face images
[8] Hariyanto, Sudiro, S. A., & Lukman, S. (2015). Minutiae Matching
and fingerprints should be taken along with the basic voter Algorithm Using Artificial Neural Network for Fingerprint Recognition.
details, such that at the time of polling a two -factor 2015 3rd International Conference on Artificial Intelligence, Modelling
authentication of the voter can be done by using Eigen -face and Simulation (AIMS). doi:10.1109/aims.2015.16
based algorithm for facial recognition and minutiae [9] RAVI. J*, K. B. RAJA**, VENUGOPAL. K. R “Fingerprint Recognition
algorithm for fingerprint recognition. The voter will be Using Minutia Score Matching” Ravi.J. et al
/International Journal of Engineering Science and Technology Vol.1(2),
allowed to cast his vote only if the eigen-faces and 2009, 35-42
fingerprints of the voter match with those in the database. [10] N. U. Ain, F. Shaukat, A.S. Nagra and G. Raja, “An Efficient Algorithm
In case of physically challenged people, who cannot submit For Fingerprint Recognition Using Minutiae” Vol. 70 No. 2 June, 2018)
their fingerprints, iris and facial recognition can be do ne.
This secure system thus avoids the vote frauds and
malpractices during elections. In future, this procedure can
be made completely digital on an online platform which can
be connected to live dashboard - making it a secure, rapid,
fraud reductant and dynamic E-voting system which
decreases a lot of mankind and resources put into the
traditional voting system.

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 248