An-ATM-with-an-eye.pdf

 ATM   ATM WITH  WITH  AN   AN  EYE  EYE

ABSTRACT

institutions alike from fraud and other breaches of security. of security.

 There is an urgent need for

improving

banking

security

region.

With

in

INTRODUCTION

the

advent of A of ATM though banking

 The rise of  technology in

became a lot easier it even

"ndia has brought into force

became a lot vulnerable.  The

many types of e$uipment of e$uipment that

chances of misuse of misuse of this of this much

aim

hyped ‘insecure’ baby product

satisfaction. ATM is one such

(ATM) are manifold due to the

machine hich made money

exponential

transactions

groth

of 

at

more

customer

easy

for

‘intelligent’ criminals day by

customers to bank.  The other

day. ATM systems today use

side of this of this improvement is the

no more than an access card

enhancement of  the culprit’s

and

probability

!"#

for

identity

verification.  This situation is

‘unauthentic’

unfortunate since tremendous

 Traditionally%  Traditionally%

progress has been made in

handled

biometric

combination

identification

techni$ues%

including

finger

by

to

get

his share.

security

is

re$uiring

the

of  a

physical

access card and a !"# or other

printing% facial recognition% and

passord in order to access a

iris scanning.

customer’s model

account.

invites

 This

fraudulent

 This paper proposes the

attempts through stolen cards%

development of  a system that

badly&chosen or automatically

integrates Facial recognition

assigned !"#s% cards ith little

and Iris scanning technology

or no encryption schemes%

into the identity verification

employees ith access to non&

process used in ATMs.  The

encrypted customer account

development of such of such a system

information and other points of 

ould

failure.

serve

consumers

www.pediain.com

to and

protect financial

1

 ATM   ATM WITH  WITH  AN   AN  EYE  EYE

'ur

paper

automatic

proposes

teller

security

model

combine

a

card%

a

machine that

physical

!"#%

an

and

ould

checking && is uni$ue to every person%

more

so

than

fingerprints.

access

electronic

2. LITERATURE REVIEW

facial recognition. y forcing ,or most of the of the past ten years%

the ATM to match a live image

the ma*ority of  ATMs used

of  a customer’s face ith an image

stored

database

that

in is

a

bank

associated

orldide ran under "M’s

no&defunct

'-/.

0oever% "M hasn’t issued a

ith the account number% the

ma*or update to the

damage to be caused by stolen

operating system in over six

cards and !"#s is effectively

years.

neutralied. 'nly hen the !"#

Movement

in

the

banking orld is no going in

matches the account and the live image and stored image match

ould

a

user

be

to directions1 Windos and 2inux. #34% a leading orld&

considered fully verified.

ide ATM manufacturer%

A system can examine *ust examine *ust

recently

the eyes% or the eyes nose and

announced

an

agreement to use Windos 5!

mouth% or ears% nose% mouth

6mbedded in its next

and eyebros% and so on.

generation "n this paper % e ill also

of 

personalied

ATMs (crmdaily.com.) Windos

look into an automatic teller

5! 6mbedded allos

machine

'6Ms to pick and choose from

providing

security the

model

customers

a

the thousands of  components

cardless% passord&free ay to

that make up Windos

get their money out of an of an ATM.

5!

 +ust step up to the camera hile your eye is scanned. The scanned. The

!rofessional%

integrated

including multimedia%

netorking and database

iris && the colored part of  the eye

the

camera

www.pediain.com

ill

be

2

 ATM WITH  AN  EYE

management

functionality.

 This makes the use of  off&the&

bulk of  transaction processing (linux.org.)  This model ould

shelf facial recognition

also ork ell for the

code more desirable because it

proposed system if  the ATMs

could easily be compiled for

processors ere not poerful

the Windos 5!

enough to $uickly

environment

and

the

perform the facial recognition

netorking and database tools

algorithms.

ill already be in place.

"n terms of the improvement of 

,or less poerful ATMs% 7A2% a

security standards% Master3ard

softare

is spearheading

development

company based in -cotland%

an effort to heighten the

provides

7alignite

encryption used at ATMs. ,or

36% hich is a modification of 

the past fe decades% many

the Windos 36 platform.

machines have used the 8ata

 This

allos

developers

that

6ncryption -tandard developed

target older machines to more

by "M in the mid

easily develop complex

9:;&bit key.

user&interaction

systems

.

86- has been shon to be

Many financial institutions are

rather easily cracked%

relying on a third choice%

hoever%

Windos #T% because of  its

computing hardare. "n recent

stability

and

maturity

as

a

given

proper

years% a ?Triple 86-@ scheme

platform.

has been put forth that uses

'n an alternative front% the

three

largest bank in the south of 

effective 9>&bit key length.

rail% anrisul% has

Master3ard no re$uires ne

installed a custom version of 

or relocated ATMs to use the

2inux in its set of to thousand

such

keys%

for

an

 Triple 86- scheme% and

ATMs% replacing legacy

by April% /<<=% both Bisa and

M-&8'- systems.  The ATMs

Master3ard ill re$uire that

send

database

re$uests

to

any ATM that supports

bank servers hich do the

www.pediain.com

3

 ATM WITH  AN  EYE

their cards must use  Triple

environment D ould all be

86-. ATM manufacturers are

controlled

no developing neer

narro limits% making hugely

models that support Triple 86-

robust

nativelyC such redesigns may

unnecessary

make them more

facial

amenable

to

also

including

ithin

certain

softare .'ne

recognition

leading algorithm

class is called

snapshot cameras and facial

image template based.  This

recognition softare% more

method attempts to capture

so

than

regards

they to

ould

be

retrofitting

in

pre&

existing machines .  There

are

proposed

global features of facial images into facial templates. #eural netorks% among other

hundreds and

of 

actual

methods% are often used to construct these templates

implementations of facial

for later matching use. An

recognition technology from all

alternative method% called

manner

geometry&based% is to explicitly

of  vendors

for

all

manner of uses. 0oever%

examine

for the model proposed in this

features of a face and the

paper% e are interested only

geometrical

in the process of facial

beteen

verification D matching a live

(Gross.) What must be taken

image to a predefined image

into

the

individual

relationship those

features

to verify a claim of  identity D not in the process of 

account% though% are certain

facial evaluation D matching a

key factors that may change

live image to any image

across live images1

in a database. ,urther% the

illumination% expression% and

environmental

pose (profile.)

conditions

under hich the verification

A

study

as

recently

takes place D the lighting% the

conducted

of 

leading

imaging

recognition algorithms% notably

system%

the

image

profile% and the processing

www.pediain.com

one

4

 ATM WITH  AN  EYE

developed by to researchers

scream profile% or s$uinted

at M"T% aback Moghaddam

eyes% dropped the recognition

and Alex !entland% and

rates significantly. 2astly%

one

a

commercial

product

matching

profile

changes

from "dentix called ,ace"t. The

orked reasonably ell hen

M"T program is based on

the initial training image(s)

!rincipal ,eature Analysis% an

ere frontal% hich alloed ;<&

adaptation of  template based


recognition. ,ace"t’s

degrees of profile

approach

uses

geometry&

changeG

hoever%

based local feature analysis.

success

oth algorithms have to be

keeping ATM users content

initialied

by

providing

the

locations of  the eyes in the

isn’t

;<&
amenable

to

ith the system.  The

natural

conclusion

to

database image% from hich

dra% then% is to take a frontal

they can create an internal

image for the bank

representation

database% and to provide a

of 

the

normalied face. "t is this

prompt to the user% verbal or

representation to hich future

otherise% to face the

live images ill be compared .

camera directly hen the ATM

"n the study% it as found that

verification process is to begin%

both

so as to avoid the

programs

handled

changes in illumination

need to account for profile

ell. This is important because

changes. With this and other

ATM use occurs day and night%

accommodations%

ith or ithout

recognition

artificial illumination. 2ikeise%

verification can rise above

the programs alloed general

:
expression changes

,ace"t’s local feature analysis

hile

maintaining

success.

0oever%

expressions% such as a

matching extreme

rates

for

method handled variations in the test cases slightly better than the !HA system used by the M"T researchers .

www.pediain.com

5

 ATM WITH  AN  EYE

Another

paper

advantages

shos

more

analysis

using

local

program that can be used on a

in

facial

verification

feature analysis systems.

variety of platforms%

,or internal representations of 

including

faces%

processors% and to determine

2,A

stores

them

embedded

topographicallyC that is% it

behavior

maintains feature relationships

match

explicitly.

 non&match cases.

 Template

based

protocols

for

the

systems% such as !HA% do not. The advantages of 2,A are

3. OUR METODOLO!" 

that analysis can be done on varying levels of ob*ect grouping%

and

that

 The first and most important analysis

step of  this pro*ect ill be to

methods can be independent

locate a poerful

of the topography. "n other

open&source facial recognition

ords% a system can examine

program

 *ust the eyes% or the eyes nose and mouth% or ears% nose% mouth and eyebros%

that

uses

local

feature analysis and that is targeted at facial verification.  This

program

should

be

and so on% and that as better

compilable on multiple

analysis algorithms are

systems% including 2inux and

developed% they can fit ithin

Windos variants% and should

the data frameork provided

be customiable to the

by 2,A

extent

 The conclusion to be dran for

of 

alloing

for

variations in processing poer

this pro*ect% then% is that facial

of the machines onto hich it

verification

ould be deployed.

softare is currently up to the

We

task of  providing high match

familiarie ourselves ith the

rates for use in ATM

internal orkings of the

transactions. What remains is

program so that e can learn

to find an appropriate open&

its strengths and limitations.

source local feature

-imple testing of this

www.pediain.com

ill

then

need

to

6

 ATM WITH  AN  EYE

program occur

ill so

evaluate

also

that its

need e

to

could

effectiveness.

ill be necessary% of  course% because e ill not have access to an actual ATM or its

-everal

softare.

sample images ill be taken of 

oth pieces of softare ill be

several individuals to be used

compiled

as test cases D one

Windos 5! and a 2inux

each

for

?account@

images%

and

run

on

a

system. 'nce they are both

and several each for ?live@

functioning properly% they ill

images% each of hich ould

be teaked as much as

vary pose% lighting conditions%

possible

and expressions.

performance (decreasing the

'nce chosen%

a

final e

program

ill

develop

increase

is

time spent matching) and to

a

decrease memory footprint.

simple ATM black box program.  This

to

,olloing that% the black boxes

program

ill

ill

be

broken

into

to

server as the theoretical ATM

components D a server

ith hich the facial

and a client D to be used in a

recognition

softare

ill

to&machine

netork.

 The

interact. "t ill take in a name

client code ill act as a user

and passord% and then

interface% passing all input

look in a folder for an image

data to the server code% hich

that is associated ith that

ill handle the calls to the

name. "t ill then take in an

facial

image from a separate folder

further reducing the memory

of  ?live@ images and use the

footprint and processor

facial recognition program

load re$uired on the client end.

to

generate

a

match

level

recognition

softare%

"n this sense% the thin client

beteen the to. ,inally it ill

architecture of many

use the match level to

ATMs ill be emulated.

decide hether or not to allo

We ill then investigate the

?access@% at hich point it ill

process of using the black box

terminate. All of this

program to control a

www.pediain.com

7

 ATM WITH  AN  EYE

I- camera attached to the

for in most cases. ,urther% a

computer to avoid the use of 

positive visual match ould

the folder of ?live@ images.

cause the live image to be

2astly% it may be possible to

stored in the database so that

add

future transactions ould have

some

sort

of 

86-

encryption to the client end to

a broader base from hich to

encrypt the input data and

compare if the original account

decrypt the output data from

image fails to provide a match

the server D knoing that

D thereby decreasing false

this ill increase the processor

negatives.

load% but better alloing us to

When a match is made ith

gauge the time it

the !"# but not the images% the

takes to process.

bank could limit transactions in a manner agreed upon by the customer hen the account as opened% and could store the image of the user for later

ATM S"STEMS

examination by bank officials. "n regards to bank employees gaining access to customer

'ur ATM system ould only attempt to match to (and

!"#s for use in fraudulent

later% a fe) discrete images%

transactions%

searching

ould likeise reduce that

through

a

large

this

system

database of  possible matching

threat to exposure to the lo

candidates

limit imposed by the bank and

ould

unnecessary.

 The

be process

agreed to by the customer on

ould effectively become an

visually

exercise in pattern matching%

transactions.

hich

ould

great

deal

not of 

re$uire time.

a

With

unverifiable

"n the case of  credit card use

at

ATMs%

such

a

appropriate lighting and robust

verification system ould not

learning

currently be feasible ithout

softare%

slight

variations could be accounted

www.pediain.com

creating an overhaul for the

8

 ATM WITH  AN  EYE

entire

credit

card

issuing

industry% but it is possible that positive

results

significant

fraud

 The bank accounts ere not

(read1 reduction)

achieved by this system might

last

(at that time) connected by a computer netork to the ATM.  Therefore% banks ere at first

motivate such an overhaul.  The

ithdran from an account.

very exclusive about ho they

consideration

is

gave ATM privileges to. Hiving

that consumers may be ary

them

of  the privacy concerns raised

holders

by

used before ATM cards) ith

maintaining

images

of 

only

to

credit

card

(credit

cards

ere

customers in a bank database%

good

encrypted or otherise% due to

modern

possible hacking attempts or

authenticate

employee

using a plastic card ith a

misuse.

0oever%

banking

records.

ATMs%

"n

customers

themselves

by

one could argue that having

magnetic

the image compromised by a

encodes

third party ould have far less

account

dire conse$uences than the

entering a numeric passcode

account

called

information

itself.

stripe% the

customerJs

number%

a

hich

!"#

and

by

(personal

,urthermore% since nearly all

identification number)% hich

ATMs

in some cases may be changed

videotape

customers

engaging in transactions% it is

using the machine. Typically% if 

no broad leap to realie that

the

banks already build an archive

incorrectly several times in a

of their customer images% even

ro% most ATMs ill retain the

if  they

card as a security precaution

are

grouped

not

necessarily

ith

account

number

is

entered

to prevent an unauthorised

information.

user from orking out the !"#

ISTOR" 

by pure guessork..

 The first ATMs ere off&line machines% as

meaning

not

www.pediain.com

money

automatically

9

 ATM WITH  AN  EYE

ARDWARE

AND

SOFTWARE

ith 2inux. 'ther platforms include 4M5 >% '-/ and Windos : bundled ith +ava.

ATMs

contain

secure

cryptoprocessors%

generally

 The neest ATMs use Windos 5! or Windos 5! embedded.

ithin an "M !3 compatible host

computer

in

a

secure

RELIABILIT" 

enclosure.  The security of  the ATMs

machine relies mostly on the integrity

of 

the

cryptoprocessor1 softare

secure

the

often

host

runs

commodity

on

a

operating

modem

over

a

via

a

dedicated

telephone line% although the move

toards

reliable% but if  they do go rong customers ill be left ithout cash until the folloing morning or henever they can

hours. 'f course% not all errors

connect directly to their ATM !rocessor

generally

get to the bank during opening

system."n&store ATMs typically

 Transaction

are

"nternet

are

to

the

detriment

of 

customersC there have been cases of  machines giving out money ithout debiting the account% or giving out higher

connections is under ay.

value notes as a result of  "n moving

addition% aay

ATMs

from

are

custom

incorrect

denomination

of 

banknote being loaded in the

circuit boards (most of  hich

money cassettes. 6rrors that

are

can occur may be mechanical

based

on

architecture)

"ntel

and

<>

into

full&

(such

as

card

transport

fledged !3s ith commodity

mechanismsC keypadsC hard

operating

disk failures)C softare (such

systems

such

as

Windos /<<< and 2inux. An

as operating systemC device

example of this is anrisul% the

driverC

largest bank in the -outh of 

communicationsC

rail% hich has replaced the

don to operator error.

application)C or

purely

M-&8'- operating systems in its automatic teller machines

www.pediain.com

10

 ATM WITH  AN  EYE

SECURIT" 

hen youJre alone% leave the 6arly ATM security

area.

focused on making the ATMs invulnerable to physical attackC

%ee& $o' (INcon)i*ential.

they

ere

Memorie

ith

dispenser

effectively

safes

mechanisms.

your

"dentification

!ersonal

#umber

(!"#)C

ATMs are placed not only near

donJt rite it on your card or

banks% but also in locations

leave it in your allet or purse.

such as malls% grocery stores%

7eep your number to yourself.

and

#ever provide your !"# over

restaurants.  The

other

side of this improvement is the

the telephone% even if  a caller

enhancement of  the culprit’s

identifies himself  as a bank

probability

employee or police officer.

to

get

his

‘unauthentic’ share.

#either person ould call you

ATMs are a $uick and

to obtain your number.

convenient ay to get cash.  They

are

also

public

and

Con*'ct

transactions

in

visible% so it pays to be careful

&ri+ate. -tay s$uarely in front

hen

of  the ATM hen completing

youJre

transactions.

making

,ollo

these

your transaction so people

general tips for your personal

aiting behind you onJt have

safety.

an opportunity to see your !"# being entered or to vie any

Sta$

"f  an

alert.

ATM

is

account information. -imilarly%

housed in an enclosed area%

fill out your depositithdraal

shut the entry door completely

slips privately.

behind you. "f  you drive up to an ATM% keep your car doors

Don,t )las- $o'r cas-. "f you

locked and an eye on your

must count your money% do it

surroundings.

at the ATM% and place your

uneasy

or

"f 

sense

you

feel

something

cash into your allet or purse

may be rong hile youJre at

before stepping aay. Avoid

an ATM% particularly at night or

making

www.pediain.com

excessively

large

11

 ATM WITH  AN  EYE

ithdraals. "f you think youJre

station

being folloed as you leave

safety information.

for

more

personal

the ATM% go to a public area near

other

people

and%

if 

FACIAL RECO!NITION

necessary% ask for help.

Sa+e

recei&t.

 Kour

ATM

receipts provide a record of 

reconcile

monthly

bank

ith

your

statement.

"f 

you notice any discrepancies on

your

statement%

developing such a model are keeping the time elapsed in

your transactions that you can later

 The main issues faced in

contact

your bank as soon as possible. 2eaving receipts at an ATM can also let others kno ho much money youJve ithdran and ho much you have in your

the verification process to a negligible amount% alloing for an

appropriate

level

of 

variation in a customer’s face hen

compared

database

image%

to

the

and

that

credit cards hich can be used at ATMs to ithdra funds are generally issued by institutions that do not have in&person

account.

contact ith the customer% and !'ar* $o'r car*. 8onJt lend your card or provide your !"# to others% or discuss your bank account

ith

friendly

strangers. "f your card is lost or stolen%

contact

your

bank

hence

no

opportunity

to

ac$uire a photo. ecause the system ould only attempt to match to (and later% a fe) discrete images% searching through a large database of  possible

immediately.

matching candidates ould be unnecessary.  The Ie*iatel$

re&ort

an$

process

ould effectively become an

crie to t-e &olice. 3ontact

exercise in pattern matching%

the

hich ould not re$uire a

8epartment

'f 

!ublic

-ecurity or your local police

www.pediain.com

great

deal

of  time.

With

12

 ATM WITH  AN  EYE

appropriate lighting and robust learning

softare%

slight

"n the case of  credit card use

at

ATMs%

such

a

variations could be accounted

verification system ould not

for in most cases. ,urther% a

currently be feasible ithout

positive visual match ould

creating an overhaul for the

cause the live image to be

entire

stored in the database so that

industry% but it is possible that

future transactions ould have

positive

results

a broader base from hich to

significant

fraud

compare if the original account

achieved by this system might

image fails to provide a match

motivate such an overhaul.

D

thereby

decreasing

credit

card

issuing

(read1 reduction)

false

negatives.

 The last consideration is that consumers may be ary of the

When a match is made ith

privacy concerns raised by

the !"# but not the images% the

maintaining

bank could limit transactions in

customers in a bank database%

a manner agreed upon by the

encrypted or otherise% due to

customer hen the account

possible hacking attempts or

as opened% and could store

employee misuse. 0oever%

the image of the user for later

one could argue that having

examination by bank officials.

the image compromised by a

"n regards to bank employees

third party ould have far less

gaining

dire conse$uences than the

!"#s

access

for

use

transactions% ould

likeise

to in

customer fraudulent

this

system

reduce

that

account

images

information

of 

itself.

,urthermore% since nearly all ATMs

videotape

customers

threat to exposure to the lo

engaging in transactions% it is

limit imposed by the bank and

no broad leap to realie that

agreed to by the customer on

banks already build an archive

visually

of their customer images% even

transactions.

www.pediain.com

unverifiable

if  they are not necessarily

13

 ATM WITH  AN  EYE

grouped

ith

account

information.

could easily be compiled for the Windos 5! environment and

the

netorking

and

database tools ill already be

SOFTWARE S(ECIFICATION

in place. ,or most of  the past ten years% the ma*ority of  ATMs used

orldide

"M’s

ran

no&defunct

Many financial institutions

under

are relying on Windos #T%

'-/.

because of  its stability and

0oever% "M hasn’t issued a

maturity as a platform.The

ma*or update to the operating

ATMs send database re$uests

system

to bank servers hich do the

in

Movement

over in

six

the

years. banking

bulk of  transaction processing

orld is no going in to

(linux.org.)  This model ould

directions1 Windos and 2inux.

also ork ell for the proposed

#34% a leading orld&ide ATM

system if the ATMs processors

manufacturer%

ere not poerful enough to

recently

announced an agreement to

$uickly

use Windos 5! 6mbedded in

recognition algorithms.

its

next

generation

personalied

Windos

choose

from

SECURIT" 

"n

the

terms

improvement

thousands of  components that

standards%

make

spearheading

up

Windos

!rofessional%

netorking

5!

including

integrated

multimedia% and

management

database functionality.

 This makes the use of  off&the&

desirable

www.pediain.com

because

of  of 

the security

Master3ard an

effort

is to

heighten the encryption used at ATMs. ,or the past fe decades% many machines have used

the

8ata

6ncryption

-tandard developed by "M in

shelf  facial recognition code more

facial

5!

6mbedded allos '6Ms to pick and

the

of  ATMs

(crmdaily.com.)

perform

the mid 9:;& it

bit key. 86- has been shon

14

 ATM WITH  AN  EYE

to be rather easily cracked%

verification D matching a live

hoever%

image to a predefined image

given

proper

computing hardare. "n recent

to verify a claim of  identity D

years% a ?Triple 86-@ scheme

not in the process of  facial

has been put forth that uses

evaluation D matching a live

three

such

keys%

for

an

image to any image in a

effective 9>&bit key length.

database.

ATM manufacturers are no

environmental

developing neer models that

under hich the verification

support  Triple 86- nativelyC

takes place D the lighting% the

such

imaging system% the image

redesigns

may

make

,urther%

the

conditions

them more amenable to also

profile% and the processing

including

environment D ould all be

and

snapshot facial

cameras

recognition

controlled

ithin

certain

softare% more so than they

narro limits% making hugely

ould

robust

be

in

regards

retrofitting

to

pre&existing

machines .

softare

unnecessary facial

.'ne

recognition

leading algorithm

class is called image template based.  This method attempts to capture global features of  FACIALRECO!NITION

facial

images

TECNI/UE0

templates.

into

What

facial

must

be

taken into account% though%  There

are

proposed

hundreds and

implementations

of 

actual of 

facial

recognition technology from all manner

of  vendors

for

are certain key factors that may

change

images1

across

live

illumination%

expression% and pose (profile.)

all

manner of  uses. 0oever% for the model proposed in this paper% e are interested only in

the

process

www.pediain.com

of 

facial

 The natural conclusion to dra% then% is to take a frontal

15

 ATM WITH  AN  EYE

image for the bank database% and to provide a prompt to the

 The

first

and

most

user% verbal or otherise% to

important step of  this pro*ect

face the camera directly hen

ill be to locate a poerful

the ATM verification process is

open&source facial recognition

to begin% so as to avoid the

program

need to account for profile

feature analysis and that is

changes. With this and other

targeted at facial verification.

accommodations%

recognition

 This

that

uses

program

local

should

rates for verification can rise

compilable

above

systems% including 2inux and

:
A

system

can

on

be

multiple

examine  *ust the eyes% or the

Windos variants% and should

eyes nose and mouth% or ears%

be customiable to the extent

nose% mouth and eyebros%

of  alloing for variations in

and so on

processing

.

poer

of 

the

machines onto hich it ould

 The conclusion to be dran

be deployed.

for this pro*ect% then% is that facial verification softare is

We

ill

then

need

to

currently up to the task of 

familiarie ourselves ith the

providing high match rates for

internal

use in ATM transactions. What

program so that e can learn

remains

its strengths and limitations.

is

to

find

an

orkings

of 

the

appropriate open&source local

-imple testing of this program

feature

ill also need to occur so that

analysis

facial

verification program that can be

used

on

a

variety

e of 

could

evaluate

its

effectiveness. -everal sample

platforms% including embedded

images ill be taken of several

processors% and to determine

individuals to be used as test

behavior

cases D one each for ?account@

protocols

for

match  non&match cases.

the

images% and several each for ?live@ images% each of  hich

OUR METODOLO!" 

www.pediain.com

16

 ATM WITH  AN  EYE

ould

vary

pose%

lighting

conditions% and expressions.

functioning properly% they ill be

teaked

possible

as

much

to

as

increase

performance (decreasing the 'nce

a

chosen%

final

e

simple

program

is

time spent matching) and to

develop

a

decrease memory footprint.

ill

ATM

program.  This

black

box

program

ill

,olloing that% the black

server as the theoretical ATM

boxes ill be broken into to

ith

components D a server and a

hich

recognition

the

facial

softare

ill

client D to be used in a to&

interact. "t ill take in a name

machine netork.  The client

and passord% and then look in

code

a folder for an image that is

interface% passing all input

associated ith that name. "t

data to the server code% hich

ill then take in an image from

ill handle the calls to the

a

facial

separate

images

folder

and

recognition generate

a

use

of  ?live@

ill

act

as

recognition

a

user

softare%

the

facial

further reducing the memory

program

to

footprint and processor load

level

re$uired on the client end. "n

match

beteen the to. ,inally it ill

this sense% the thin client

use the match level to decide

architecture of many ATMs ill

hether

be emulated.

or

not

to

allo

?access@% at hich point it ill terminate. All of  this ill be

We ill then investigate the

necessary% of  course% because

process of using the black box

e ill not have access to an

program to control a I-

actual ATM or its softare.

camera

attached

to

the

computer to avoid the use of  oth pieces of softare ill

the folder of  ?live@ images.

be compiled and run on a

2astly% it may be possible to

Windos

add

5!

and

a

2inux

system. 'nce they are both

www.pediain.com

some

sort

of 

86-

encryption to the client end to

17

 ATM WITH  AN  EYE

encrypt the input data and

remember

decrypt the output data from

identification

the server D knoing that this

passord and scratch it on the

ill

back

increase

the

processor

of 

their

personal

number

their

cards

or

or

load% but better alloing us to

somehere that a potential

gauge the time it takes to

thief can find% no more fear of 

process.

having an account cleaned out if the card is lost or stolen.

IRIS RECO!NITION0 OW TE S"STEM WOR%S. "nspite security

of 

all

features%

technology

these a

ne

has

developed.

ank

been Inited

of 

 Texas became the first in the Inited

-tates

recognition automatic providing

to

iris

technology

at

the

a bankcard% a stereo camera locates the face% finds the eye and takes a digital image of 

offer

teller

When a customer puts in

machines%

customers

a

the iris at a distance of  up to three

feet.

 The

resulting

computeried Liris codeL is compared

ith

one

the

cardless% passord&free ay to

customer ill initially provide

get their money out of an ATM.

the bank. The ATM onJt ork

 ThereJs

no

card

to

sho%

thereJs no fingers to ink% no customer

inconvenience

discomfort.

"tJs

 The entire process takes less

or

 *ust

customerJs eyes.  +ust step up to the camera hile your eye && the

colored part of  the eye the camera ill be checking && is uni$ue to every person% more so than fingerprints. And% for the

customers

www.pediain.com

ho

than to seconds.

a

photograph of  a ank Inited

is scanned.  The iris

if  the to codes donJt match.

canJt

 The

system

orks

e$ually ell ith customers earing glasses or contact lenses and at night. #o special lighting is needed. The camera also does not use any kind of  beam. "nstead% a special lens has been developed that ill not only blo up the image of 

18

 ATM WITH  AN  EYE

the

iris%

but

provide

more

When the system is fully

detail hen it does. "ris scans

operational% a bank customer

are much more accurate than

ill have an iris record made

other

for

high&tech

available

that

"8

systems

scan

voices%

faces and fingerprints.

comparison

hen

an

account is opened.  The bank ill

have

the

option

of 

identifying either the left or -cientists

have

identified

/=< features uni$ue to each personJs iris && compared ith about F< for fingerprints && and it remains constant through a personJs life% unlike a voice or a face. ,ingerprint and hand patterns

can

be

changed

through alteration or in*ury.  The iris is the best part of the eye

to

use

as

a

identifier

because there are no knon diseases of  the iris and eye

right eye or both. "t re$uires no intervention by the customer.  They ill simply get a letter telling them they no longer have to use the !"# number. And% scam artists beare% a picture of  the card holder onJt pass muster.  The first thing the camera ill check is hether the eye is pulsating. "f  e donJt see blood floing through your eye% youJre either dead or itJs a picture.

surgery is not performed on the iris. "ris identification is the most secure% robust and stable form of identification knon to man. "t is far safer% faster% more secure and accurate than 8#A

testing.

6ven

identical

tins do not have identical irises.  The

iris

remains the

same from 9 months after birth until five minutes after death.

www.pediain.com

firstdata.com 4etail ATM !rogram  Triton 42=<<< Model  The  Triton automatic

42=<<< teller

model machine

(ATM) brings a superior level of  performance% reliability and customiation options to the embedded !3 ATM industry.

19

 ATM WITH  AN  EYE

 This

durable

ATM

solution

installation%

comprehensive

allos merchants to offer their

training% technical support%

customers the convenience of 

service and reporting features.

an on&site ATM% hich may

42=<<<

lead

firstdata.com

to

increased

customer

traffic% increased sales volume

N<9&NNN

and

4etail ATM !rogram

an

additional

revenue

source. ,rom small specialty

 Triton 42=<<< Model

stores to larger petroleum and

Feat'res

convenience stores% the

 ++Monitor

 Triton 42=<<< ATM is easily

& 9<.FL BHA color screen

configured to meet the needs

 ++Advertising !otential

of merchants of all sies.  The

 Triton

& 8ecals and signage

42=<<<

ATM

features the user&friendly

& 3ouponing & 8isplay graphics

Microsoft Windos 36 =.<

 ++!rinter

operating system and has a

& 
large

graphics capabilities

capacity

for

storing

 *ournal records or donloading graphics

for

marketing

improved

efforts.

 The

prominent

profile make the ATM an

other

& !3" compliant

component

on&site

& -tandard dial  ++-mall ,ootprint

of 

any merchant location. Many

& 6ncrypting !"# !ad (6!!)

 ++3ommunications

9<.FL 238 display and sleek

eye&catching

 ++7eypad

 ++acklit Topper &

ATM

programs can be expensive and difficult to implement% hile

Mid&topper

standard

increases visibility  ++3ard 4eader & 8ip standard  ++A8A 3ompliant

the Triton 42=<<< ATM offered

& Americans ith 8isabilities

by ,irst 8ata is a complete

Act

ATM solution that includes

www.pediain.com

20

 ATM WITH  AN  EYE

(A8A)

compliant

and

audio

compliant

4elative humidity1 /
 ++Technical -pecs

!oer consumption1 /.
& 4"-3 microprocessor

99= BA3 at >< 0C

& Microsoft W"ndos 36 =.<

9.
&

A

9/

M

flash

memory%

!lo1al

Lea*er

in

expandable

Electronic Coerce

& 9/ M 4AM

,irst 8ata poers the global

& Multilingual capabilities

economy by making it easy%

& 3assette options available

fast and secure for people and

 ++-ecurity

businesses around the

& 6lectronic lock standard

orld

Bene)its

services using virtually any

 ++-upplementary

revenue

source (potential profit from

ATM

buy

goods

and

form of payment. -erving millions of 

surcharge

andor

interchange)  ++0igher

to

merchant locations and thousands of  card issuers% e

sales

volume

from

more in&store purchases

have the expertise and insight to help you accelerate

 ++"ncreased store traffic

your business. !ut our

 ++-uperior

intelligence to ork for you.

customer

convenience

,or more information% contact

 ++"mproved customer loyalty

your

Diensions

,irst 8ata -ales 4epresentative

0eight1 >F.:L (ith mid&topper)

or visit firstdata.com.

Width1 9>.:L

Q /<<: ,irst 8ata 3orporation.

8epth1 /=.>L

All

Weight1 /F= lbs.

trademarks% service

3olor1 ayou rone

marks

O&erating En+ironent

referenced in this material are

 Temperature

range1

F
www.pediain.com

9
to

rights

and

reserved.

trade

All

names

the property of their respective oners.

21

 ATM WITH  AN  EYE

CONCLUSION0

one%

Mike%

 +ames We thus develop an ATM

2.%

Wayman% and

8uane.

8r.

lackburn% ?6valuating

model that is more reliable in

,acial4ecognition  Technology

providing

for 8rug 3ontrol Applications.@

security

by

using

facial recognition softare. y

'#83!

"nternational

keeping the time elapsed in

3ounterdrug

the verification process to a

-ymposium1 ,acial 4ecognition

negligible amount e even try

Bendor Test. 8epartment of 

to maintain the efficiency of 

8efense

this ATM system to a greater degree. 'ne could argue that having

the

 Technology

3ounterdrug

 Technology

8evelopment

!rogram 'ffice% +une /<<9.

image

compromised by a third party

Hross% 4alph% -hi%  +ianbo% and

ould

3ohn%  +effrey ,. ?Suo vadis

have

far

conse$uences account

less than

information

dire the itself.

,ace

4ecognition.@

 ThirdWorkshop on 6mpirical

,urthermore% since nearly all

6valuation

ATMs

3omputer

videotape

customers

engaging in transactions% it is

Methods Bision.

in 7auai1

8ecember /<<9.

no broad leap to realie that banks already build an archive of their customer images% even if  they

are

grouped

not

necessarily

ith

account

information.

!enev% !enio -.% and Atick%  +oseph

 +.

Analysis1

?2ocal A

-tatisticalTheory

Heneral for

4epresentation.@ 3omputation

BIBILO!RA("0

,eature

'b*ect

#etork1 in

#eural

-ystems% Bol. ;% #o. N% pp. F;;&=<<% 9::>.

All% Anne. ?Triple 86- dare you.@

ATM

9: Apr. /<
www.pediain.com

Marketplace.com.

Wrolstad% +ay. ?#34 To 8eploy #e Microsoft '- in ATMs.@

22

 ATM WITH  AN  EYE

34M8aily8ot3om.

/:#ov.

/<<9

www.pediain.com

23