International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882

Volume 4, Issue 8, August 2015

SECURING ATM BY IMAGE PROCESSING FACIAL

RECOGNITION AUTHENTICATION
T. SUGANYA1, T. NITHYA2
C. SUNITHA3 , B. MEENA PREETHI4
1, 2
III M.Sc (software systems)
3
Head of Department of computer Application and Software Systems,
4
Assistant Professor, Department of computer Application and Software Systems
SRI KRISHNA ARTS AND SCIENCE COLLEGE
An autonomous college Affiliated to Bharathiar University
Accredited by NAAC with A Grade, An ISO 9001:2008 Certified Institution
KUNIAMUTHUR, COIMBATORE TAMILNADU - INDIA

ABSTRACT
With the support of ATM, though banking becomes easier,
it also became feeble. There has been countless gear of
abuse that have in use in banking transactions. Thus there
is a essential need to provide high security. This paper
proposes the amalgamation of Face Recognition System in
the identity verification process engaged in ATMs to
enhance the security system.
Keywords: ATM System, Face Recognition Software
(FRS), Security.

I.

INTRODUCTION

The rise of technology bring into force loads of types of

tools that aspire at more customer pleasure. ATM is a
machine which made money transactions effortless for
customers. But it has both advantages and disadvantages.
Current ATMs make use of naught more than an access
card and PIN for uniqueness confirmation. This has ATM
Using Face Recognition System demonstrate the way to a
lot of fake attempt and mistreatment through card theft,
PIN theft, stealing and hacking of customers account
details and other part of security.

II.

WHAT ARE
SYSTEMS?

FACE

RECOGNITION

FRS is an application that mechanically identifies a person

from a digital image or a video outline from a video
source. One of the behaviors to do this method is by
matching chosen facial features from a facial database and
the image.

III.

WHEN DID THEY DEVELOP?

The pioneer of robotic Face Recognition includes Helen

Chan Wolf, Woody Bledsoe and Charles Bisson. During
1964 and 1965, Bledsoe, all along with Helen Chan Wolf
and Bisson, worked on using the computer to be familiar

with human faces. He was pompous of his work, but

because the support was provided by an unknown
intelligence group that did not allow more publicity, slight
work was published. Given a large database of photos and
images, the trouble occurred was to select from the
database a tiny set of records such that one of the image
records coordinated the photograph. The achievement of
the performance could be well thought-out in terms of the
relation of the respond list to the quantity of report in the
database. Still the recognition problem was made hard by
the great discrepancy in lean and head rotation, lighting
intensity and angle, facial expression, etc. In 1966, the
work was continued first and foremost by Peter Hart.
Peters experiment done on a database contains over 2000
images, the computer eternally outperformed humans
when presented with the same detection tasks. The
enlargement period for facial recognition started in the
delayed 1980s and they were existing systems was made
accessible in the 1990s.

IV.

HOW DO THEY WORK?

A database of peoples face is maintained by the system

that handles face detection. When a face needs to be
predictable a snap of the ones look is taken and evaluated
to the appearance present in the database to observe if a
match is found. There are typically 3 parts related to a face
recognition system
Face detector,
Eye localizer and
Face recognizer
The face detector:
The face detector spot the face, eliminating any other
detail, not related to the face (like the backdrop). It
identifies the facial region and leaves the non-facial region
in the photo of the person to be identified.

www.ijsret.org

913

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 8, August 2015

The eye-localizer:
It finds the spot of the eyes, so that the position of the face
can be identified better.
The recognizer:
It will checks the database to find a match.

V.

ALGORITHM
FOR
RECOGNITION SYSTEMS

FACE

Various facial recognition algorithms be familiar with

faces by extracting features, from a snap of the subject's
face. For example, an algorithm may examine the size,
relative position, in addition to/or outline of the nose, eyes,
cheekbone and jaw. These facial appearances are then used
to search for other imagery across matching features.
Other algorithm manage a balcony of face images and then
compress the images face information and it saves only
the data in the image that is used for face detection. A
searched image is then compared with the face record.
Face Recognition Systems algorithms is alienated into two
main approaches namely
Geometric based
Image-template based.

VI.

appearance like contours of the eye sockets, chin, nose,

peaks and valley on the visage for identification. The
database will store details of faces also. The advantage of
3-D technique over 2-D method is that 3-D face
identification works fine even if the face is turned at 90
degree to the camera. It is self-governing of lighting
environment and facial expressions.

TECHNIQUES AND METHODS

Surface Texture Analysis

The most superior method is Surface Texture Analysis
(STA). STA does not examine the entire face but a patch
of membrane on it. This patch is divided into separate
blocks. The skin surface, the pore on the skin and other
face characteristics are converted to a code. This code is
used for comparison.

VII.

They are of three types

2-D,
3-D
Surface Texture Analysis.
2-D Technique
The 2-D recognition method was individual of the original
techniques employed. It maintained details of peoples
faces as seen two dimensionally. Details like width of the
nose, width of the eyes, distance between the eyes, jaw
line, cheek bone figure were used for contrast. This type of
face recognition was not too precise. Change in facial
expression or difference in ambient lighting on a
appearance that is not directly looking into the camera did
not produce expected results.

TELLER

For the past ten years, the bulk of ATMs used worldwide
ran under IBMs now out-of-date OS/2. Movement in the
banking world is now going on using Windows and Linux.
A leading world-wide ATM manufacturer, NCR currently
announced an harmony to employ Windows XP
Embedded in its next generation of personalized ATMs.
For fewer influential ATMs, KAL, a software
development company based in Scotland, gives Kalignite
CE, which is a modified version of Windows CE platform.
This platform allows developers that target older machines
to more easily build up multifaceted user-interactive
systems. Many financial institutions are having a third
choice, Windows NT. It is because of its maturity and
stability as a platform. On an alternative, the largest bank
in the south of Brazil, Banrisul, had installed a custom
version of Linux in its put of two thousand ATMs,
replaces legacy MS-DOS systems. In these types of
ATMs, it sends database requests to bank servers which do
the bulk of transaction processing.

VIII.
3-D Technique
Progression in face recognition gave origin to the 3-D
recognition system. This stepped up technique, used facial

ATM
(AUTOMATIC
MACHINE)

WORKING OF FRS IN ATM

Face Recognition Systems (FRS) labor in ATMs in the

following way
Initially the customers picture is taken when the
account is open and the user is allowed to set nonverified transaction limits.

www.ijsret.org

914

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 8, August 2015

At ATM, access card and PIN are used to preverify user.

Users snap is taken and an attempt is made to
match it to the record image.
If the match procedure becomes successful, allow
transaction.
If the match is unsuccessful, limit the available
transactions.
When a match is complete with the PIN but not the
imagery, the bank could limit the transactions in a way
granted upon by the user when the account was opened,
and could store the photograph of the user for later
examination by bank official. In the case of using credit
card at ATMs, confirmation system would not presently be
feasible without creating a repair for the entire credit card
issuing industry.

Working of FRS in ATM

IX.

METHODOLOGY

The important step is to locate a dominant open-source

appearance identification program that uses local feature
analysis and that is based on facial verification. This plan
must be compliable on multiple systems, involving
Windows and Linux variants, and also be customizable to
the degree of allowing for variations in processing power
of the machines onto which it would be deployed.
Once a final program is done, a simple ATM black box
program is developed. This program will serve as the

hypothetical ATM with which the facial recognition

software will work together. It will take a name and
password, and then look into a folder for a photograph that
is associated with that name. It will then take an image
from a separate folder of live images and use the facial
recognition program to produce a match level between the
two images. At last it will use the match level to decide
whether access or not to allow access, at which point
it will terminate. Both pieces of software will be checked
for errors and run on a Windows XP and a Linux system.
If both are functioning properly, they will be tweaked as
much as potential to increase performance (decreasing the
time used up matching) and to decrease memory footstep.

Following this, the black boxes will be broken into two

parts a server and a client It is done to use in a network
consist of two-machine. The client code will act as a user
interface, which is passing all input data to the server code.
It will handle the calls to the facial recognition software
(FRS), further reducing the buffer footstep and processor
load essential on the customer end. The black box program
is used to control a USB camera which is attached to the
computer. It is done to avoid the use of the folder of live
images. At last, it may be possible to add some sort of
encryption to the client end to encrypt the input and
decrypt the output from the server knowing that it will
boost the processor load, but better allowing us to measure
the time it takes for processing.

www.ijsret.org

915

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 8, August 2015

X.

KEY FACTORS TO BE CONSIDERED

There are certain factors that affect process of support.

They are
1. Lighting
2. Extreme facial expressions
3. Angle of view
4. Facial hair
5. Glasses
Other concerns are keeping the time beyond in the
verification process to a small amount, allowing for an
suitable level of variation in a customers face when
compared to the database photograph, and the credit cards
that can be used at ATMs to withdraw funds are generally
provided by institutions, these institutions do not have
personally contact with the customer, and hence no chance
to acquire a photo. The last concern is that consumers may
be wary of the privacy concerns raised by maintaining
photographs of customers in a bank record, encrypted
orelse, due to possible hacking attempts or employee
abuse.

XI.

CAN ALL ATMs SUPPORT FRS?

Most recent generation ATMs support Windows CE, 2000,

XP Embedded, or Linux these machines can also support
facial recognition software locally elder ATMs run DOS or
OS/2 these machines can offload the processing to the
banks computers. As we have seen before both Local
Feature Analysis (LFA) and Principle Feature Analysis
(PFA) programs have handle changes in lighting well. This
is significant because ATM is active every 24 hours, with
or without artificial illumination. Verification rates can be
as elevated as 90% when the factors that affect facial
recognition systems are taken care of.

XII.

ADVANTAGES AND DISADVANTAGES

Advantages
Verification rates as high as 90% have been
attained when face recognition system had used in
ATMs.
It has been used to strengthen security.
It can be used to reduce fraudulent attempts.
The procedure used in Face Recognition Systems
handle the changes in the light effectively. This is
important since ATM use occur day and night,
with or without fake light.
With appropriate lighting and strong learning
software, slight variations in the images could be
accounted for.
Positive visual match would cause the existing
picture to be stored in the record so that future
transactions would have a broader foot from which

to compare if the original account photograph fails

to provide a match.
When a match is complete with the PIN but not
the imagery, the bank could limit the transactions
in a way granted upon by the user when the
account was opened, and could store the
photograph of the client for later examination by
bank official.
In regards to bank staff gaining access to customer
PINs for use in fraudulent transactions, this system
will reduce the threat to contact to the low limit
forced by the bank and agreed to by the user on
visually unverifiable transactions.
Disadvantages
Not identifying people correctly even if their
photo is in the database. Changes in lighting and
expressions like scream expressions, squinted
eyes, changes in disguise like wearing hats;
glasses drop recognition rates significantly even
though the user is a genuine account holder.
Matching profile changes worked reasonably well
when the first guidance image(s) were frontal,
which allowed 70-80% success rates for up to 45
degrees of profile change however, 70-80%
achievement isnt amenable to keeping ATM users
content with the system.
Consumers may be cautious of privacy concerns
raised by maintaining images of clients in a bank
database, encrypted or else, due to feasible
hacking attempts or employee abuse.

XIII.

CONCLUSION

Facial verification software is at present up to the task of

provided that important match rates for use in ATM
transactions. Adding up facial recognition systems to the
identity confirmation process used in ATMs can reduce
forged transactions to a great extent.

REFERENCES
[1] Penev and Atick, Joseph J. Local Feature
Analysis: A
General Statistical Theory for ject Representation. Network:
Computation in Neural Systems, Vol. 7, No. 3, pp. 477-500,
1996.
[2] Gross, Ralph, Shi, Jianbo, and Cohn, Jeffrey F. Quo vadis
Face Recognition. Third Workshop on Empirical Evaluation
Methods in Computer Vision. Kauai: December 2001
[3 ]Bone, Mike, Wayman, Dr. James L., and Blackburn, Duane.
Evaluating Facial Recognition Technology for Drug Control
Applications. ONDCP International Counterdrug Technology
Symposium: Facial Recognition Vendor Test., June 2001.
[4] S. Sruthy. "Literature Survey Automated Person Identification Techniques." (2013).

www.ijsret.org

916