Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
23
,
No.
1
,
Ju
ly
2021
, p
p.
5
29
~
5
39
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v
23
.i
1
.
pp
5
2
9
-
5
39
529
Journ
al h
om
e
page
:
http:
//
ij
eecs.i
aesc
or
e.c
om
Pay
l
oad
and
qu
ality augm
entation
usin
g s
t
egan
ogra
ph
ic
optimiz
ation t
ec
h
niq
ue base
d
on edge d
ete
ctio
n
Maf
az
A
lanez
i, Iman S
ubhi
Moham
med
A
lta
ay, Saja
Y
ouni
s H
amid
Ma
ll
a'
aloo
Depa
rtment
o
f
C
om
pute
r
Scie
n
ce,
Col
le
ge
of
Com
pute
r
Scie
n
ce
and
Math
ematic
s,
Univer
sit
y
of M
osul,
Ira
q
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Oct
1
8
, 202
0
Re
vised Ju
n
7
,
2021
Accepte
d
J
un
1
4
, 202
1
Inform
at
ion
sec
urity
is
on
e
of
t
he
m
ost
signifi
c
ant
pro
ce
ss
es
th
at
m
ust
b
e
ta
ken
int
o
a
cc
ou
nt
when
conf
identia
l
l
y
tra
nsferr
i
ng
informati
on.
Thi
s
pape
r
int
roduc
es
a
ste
ganogr
aph
y
t
echnique
using
th
e
edge
d
etec
t
io
n
m
et
hod.
It
foc
used
on
three
basic
and
impor
ta
nt
asp
ec
ts’
pa
yloa
d,
qu
al
i
t
y
,
an
d
sec
uri
t
y
.
W
el
l
-
known
edge
det
e
ct
ors
we
re
used
to
generat
e
as
m
an
y
ed
ge
pixe
ls
as
poss
ibl
e
to
h
ide
dat
a
and
ac
h
ie
v
e
the
highe
st
pa
y
lo
ad.
The
least
signifi
c
ant
bit
(
LSB
)
al
gorithm
has
bee
n
im
prove
d
b
y
ex
te
n
ding
the
bit
s
us
ed
to
embed
bet
wee
n
2
-
4
bi
ts
in
sm
ooth
a
nd
sharp
are
as.
To
inc
re
ase
sec
uri
t
y
,
t
h
e
tra
nsac
ti
on
be
twee
n
the
two
par
ti
es
is
base
d
on
divi
ding
the
k
e
y
and
the
cove
r
imag
e
into
seve
ral
par
ts
and
agr
e
ei
ng
on
the
t
y
p
e
of
edg
e
detec
ti
on
.
The
exp
eri
m
ent
s
ac
hi
eve
d
t
he
m
axi
m
um
loa
d,
fo
r
insta
nc
e
with
a
fuz
z
y
edge
det
e
ct
or,
at
f
irst,
embedding
in
4
bit
pla
n
es
if
edg
e
pixel,
and
in
2
bit
pla
n
es
if
non
-
edge
pixel,
the
pea
k
signa
l
-
to
-
noise
rat
io
(
PSNR
)
inc
rea
sed
fr
om
43.
580
to
45.
790
.
At
s
ec
ond,
embeddi
ng
in
2
bi
tpl
an
es
if
edg
e
pix
e
l,
and
in
4
bit
pla
n
es
if
non
-
edge
pix
el,
th
e
P
SN
R
dec
rea
sed
bet
wee
n
38.
433
-
41.
593.
The
suggested
sche
m
e
a
chi
ev
ed
a
hig
h
pa
y
lo
ad
to
embed
in
the
cove
r
image
and
ac
cor
d
ing
to
hu
m
an
per
c
ept
ion
,
it
pr
ese
rve
d
the
nat
ure
of the
ori
gina
l
image
.
Ke
yw
or
ds:
Cov
e
r
im
age
Ed
ge dete
ct
ion
Payl
oad
Stegan
ogra
phic
Stego i
m
age
Qu
al
it
y
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Ma
faz A
la
nezi
Dep
a
rtm
ent o
f C
om
pu
te
r
Scie
nce
Coll
ege
of
C
om
pu
te
r
Scie
nc
e an
d
Ma
them
at
ic
s
Un
i
ver
sit
y o
f M
os
ul,
Iraq
Em
a
il
:
m
afaz
m
halanezi@
uom
os
ul
.edu.i
q
1.
INTROD
U
CTION
In
to
day'
s
world
an
d
in
th
e
tim
e
of
co
rona
virus
dise
ase
2019
(
C
OVID
-
19
)
,
f
ul
l
el
ect
ro
nic
com
m
un
ic
at
ion
has
bec
om
e
a
li
feli
ne
to
co
ntinu
e
w
ork,
e
du
cat
io
n,
resea
rch
act
ivit
ie
s,
and
oth
e
r
im
po
rta
nt
li
fe
re
qu
ir
em
e
nts.
The
refor
e
,
data
is
tra
ns
fe
rr
e
d
from
on
e
par
ty
to
ano
t
he
r
us
in
g
va
rio
us
app
li
cat
ion
s
s
uch
as
e
m
ai
l,
so
ci
al
m
edia,
an
d
oth
e
r
s.
It
has
bec
ome
incre
asi
ngly
i
m
po
rtant
to
protect
our
pri
va
te
inform
ation
from
m
isuse
by
at
tack
ers
.
The
pro
cess
of
e
xch
a
ngin
g
co
nf
i
den
t
ia
l
and
i
m
po
rt
ant
inf
or
m
at
ion
over
netw
orks
is
a
gr
eat
m
otivati
on
for
researc
he
rs
inte
rested
in
fin
ding
a
n
ap
proac
h
t
o
c
reate
a
sec
ur
e
pat
h
fo
r
this
in
form
at
ion
away
f
ro
m
sp
yi
ng
a
nd
pe
netr
at
ion
.
In
this
s
ense,
va
rio
us
e
ncr
y
ption
al
gorithm
s
hav
e
be
en
create
d
to
c
onve
r
t
this
inf
or
m
at
ion
int
o
un
read
a
ble
data
durin
g
the
t
ran
sm
issi
on
proces
s
over
netw
orks
.
W
it
h
the
de
vel
op
m
ent
of
e
ncr
y
pt
re
ve
rse
en
gi
nee
ri
ng
m
et
ho
ds
an
d
enc
rypt
a
naly
sis,
weakness
es
ha
ve
arise
n
in
these
e
ncry
ption
al
gorithm
s,
an
d
it
has
bec
om
e
po
ssi
ble
t
o
pen
et
rate,
re
tur
n,
a
nd
read
enc
rypted
dat
a.
T
hat
is
why
these
m
et
ho
ds
hav
e l
os
t t
heir
pri
vac
y and
secu
rity
to
so
m
e extent. Th
is prom
p
te
d
the r
esearc
hers to search
fo
r
oth
e
r
m
et
ho
ds
to
obta
in
m
or
e
co
nfi
den
ti
al
it
y.
Ther
ef
or
e
,
m
e
thods
ha
ve
bee
n
us
e
d
to
hid
e
i
nfor
m
at
ion
in
dig
it
al
m
edia
that
can
be
a
n
im
age,
vi
deo
,
or
te
xt
wi
thout
af
fecti
ng
the
qual
it
y
of
these
m
edia
an
d
in
a
way
that
does
no
t
cause
network
hack
e
rs
t
o
s
uspect
the
existe
nce
of
th
is
inf
or
m
at
ion
within
them
,
thu
s
e
ns
uri
ng
t
hat
th
e
netw
ork d
oes n
ot interce
pt a
nd
pen
et
rate t
his
d
at
a.
This is
c
al
le
d
Stega
nograp
hy tech
nolo
gy [1]
,
[2
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
5
2
9
-
5
3
9
530
The
re
a
re
m
any
al
go
rithm
s
use
d
in
s
te
ga
nogr
a
phy
.
O
ne
of
the
sim
plest
and
m
os
t
com
m
on
of
the
se
m
et
ho
ds
is
th
e
LSB
m
et
hod,
w
hich
de
pe
nd
s
on
the
in
cl
us
io
n
of
eve
ry
bit
of
the
m
essage
in
th
e
le
ast
sign
ific
a
nt
bit
of
eac
h
plain
pix
el
in
the
i
m
age
[3
]
.
T
he
le
ast
sign
ific
ant
bit
m
e
tho
d
hid
es
in
f
or
m
ation
by
em
bed
di
ng
it
i
n
t
he
c
ov
e
r
im
age
bits
se
qu
e
ntial
ly
,
m
aking
it
easi
er
f
or
an
at
ta
cke
r
t
o
get
a
secret
m
essag
e
wh
e
n
it
i
nterce
pts.
T
o
increa
s
e
the
c
om
plexity
and
safety
of
this
m
et
hod,
it
can
be
c
om
bin
ed
with
the
edg
e
detect
ion
al
go
r
it
h
m
to
place
the
m
essage
in
the
i
m
age
de
pendin
g
on
t
he
edg
e
,
so
t
he
m
essage
is
rando
m
l
y
distrib
uted.
T
hi
s
giv
es
ad
diti
on
al
secur
it
y
for
the
hid
e
opera
ti
on
as
well
as
increase
d
stora
ge
capaci
ty
[4
]
,
[5
]
.
In
a
dd
it
io
n
to
confide
ntial
it
y,
there
is
ano
t
he
r
i
m
po
rta
nt
conditi
on
t
hat,
if
fu
l
fill
ed,
m
a
kes
the
co
ncea
l
m
ent
schem
e
su
cces
sfu
l.
T
his
co
nd
it
ion
is
a
n
inc
r
eased
ca
pacit
y
without
si
de
e
f
fects.
T
his
c
on
diti
on
has
bec
om
e
a
researc
h
c
halle
ng
e
t
hat
has
ta
ken
m
ulti
ple
f
or
m
s
to
show
m
any
prov
e
n
a
nd
qual
it
y
al
go
rithm
s.
One
way
to
increase
y
ou
r
load
is
t
o
us
e
ed
ge
detect
ion
on
t
he
c
ove
r
ph
oto
[
6]
,
[
7]
.
This
is
beca
us
e
the
ed
ge
r
egio
n
tolerat
es
chang
es
in
pix
el
values
an
d
m
essa
ges
can
be
co
m
bin
ed
with
a
gr
eat
er
capaci
ty
than
s
m
oo
th
sp
ace.
Ed
ge
detect
ion
with
var
io
us
e
dg
e
detect
or
s
l
ike
cann
y,
s
o
be
l,
and
pre
witt
.
Also
,
hy
br
i
d
de
te
ct
or
s
can
be
us
ed
to inc
rease t
he a
rea
of
t
he
e
dg
e re
gion,
t
her
e
by inc
reasin
g
t
he
m
essage loa
d
that
will
b
e
inclu
ded in it
[8]
,
[9
]
.
In
this
w
ork,
we
de
pend
on
i
m
age
ste
gan
ogra
phy,
wh
ic
h
us
es
i
m
ages
as
the
cov
e
r
file
to
h
ide
the
secret
m
essage
beca
us
e
t
hey
con
ta
in
a
l
ot
of
re
dundancy,
a
nd
re
dunda
ncy
inf
or
m
at
ion
is
the
bits
of
a
n
obj
ect
that
prov
i
de
accuracy
fa
r
gr
e
at
er
than
neces
sary
for
the
ob
j
ect
’s
use
an
d
disp
la
y,
the
be
nef
it
of
the
se
bits
is
that
they
can
be
al
te
red
wi
th
out
the
al
te
rati
on
bei
ng
d
et
ect
ed
easi
ly
[7
]
.
Th
e
portable
net
work
g
ra
phic
s
(
PNG
)
form
at
was
use
d
f
or
gray
sca
le
and
c
olou
re
d
im
ages,
as
this
f
or
m
at
was
sp
eci
al
ly
design
e
d
for
tra
ns
f
err
in
g
i
m
ages
over
the
i
nte
rn
et
,
P
NG
al
so
sup
ports
i
nd
e
xe
d
c
olor,
gr
ay
sca
le
an
d
red
,
gr
ee
n
,
a
nd
blu
e
(
R
GB
)
,
i
n
add
it
io
n
to
it
su
pp
or
ts
c
olor
palet
te
-
base
d
i
m
ages
from
R
GB
i
m
ages
of
24
-
bit
or
re
d
,
gr
ee
n
,
blu
e
,
an
d
al
ph
a
(
RGBA
)
32
-
bit.
N
on
-
c
olor
plate
-
base
d,
gray
scal
e,
an
d
fu
ll
-
colo
r
RGB
im
ages
[10].
We
hav
e
al
so
w
ork
ed
on
a
com
b
inati
on
of
LSB
s
ubsti
tuti
on
an
d
e
dge
detect
io
n
m
echan
ism
s,
s
om
e
of
wh
ic
h
are
base
d
on
a
sing
le
der
i
vative
s
uc
h
as
robe
rt,
s
obel
,
pr
e
witt
,
and
can
ny,
or
a
seco
nd
de
riva
ti
ve
su
c
h
as
la
placi
an
of
ga
ussi
an
(LOG). I
n
a
dd
i
ti
on
to
the
f
uzz
y l
og
ic
that inc
reases e
dg
e e
xpan
sio
n,
a
hybri
d detec
tor ha
s also b
ee
n use
d.
T
he
div
e
rsity
of
th
e
reag
e
nts
us
e
d
ai
m
s
to
crea
te
a
la
rg
er
inc
orp
or
at
io
n
are
a.
A
com
par
is
on
has
been
m
ade
to
choose
t
he
bes
t
appr
oach
f
or
achievin
g
high
qual
it
y
based
on
PS
NR,
m
ean
s
qu
a
re
e
rror
(
MSE
)
,
a
n
d
bit
s
pe
r
pix
el
(BPP)
qu
al
it
y
m
et
rics. Color im
ages in
the wor
k we
re
chosen t
o
inc
re
ase capaci
ty
and em
bed
di
ng in 24
-
bit
rathe
r
tha
n
8
-
bit
as
in
gr
a
ysc
al
e
i
m
ages.
The
stre
ngth
i
n
the
pro
posed
syst
e
m
co
m
es
from
achievin
g
hi
gh
secrecy
us
in
g
a
key
sp
li
t
int
o
seve
ral
par
ts
.
In
a
gray
i
m
age,
the
co
ve
r
i
m
age
is
sp
li
t
into
blo
c
ks
and
th
e
arr
a
ng
em
ent
of
these
blo
c
ks
is
the
first
par
t
of
the
key
.
Also
,
the
te
xt
is
div
ided
into
blo
c
ks
an
d
the
arr
a
ng
em
ent
of
these
blo
c
ks
i
s
the
oth
e
r
pa
rt
of
the
key.
As
for
t
he
col
or
e
d
im
ages,
the
key
has
bee
n
di
vid
ed
into
three
pa
rts,
m
eaning
the
re
is
the
thir
d
par
t
in
a
dd
it
io
n
to
the
tw
o
pa
rts
of
t
he
key
us
ed
i
n
gray
im
ages,
wh
ic
h
is
div
idi
ng
the
te
xt
into
three
pa
rts,
an
d
each
sect
io
n
is
the
siz
e
of
th
e
cov
e
r
i
m
age.
The
key
is
us
e
d
f
or
te
xt
retrieval
wh
e
n
the
dec
odin
g
is
blin
d.
The
rem
ai
nd
er
of
t
he
pa
per
has
be
en
orga
nized
as
t
he
f
ol
lowing:
Sect
ion
2
c
on
t
ai
ns
a
br
ie
f
de
scriptio
n
of
th
e
pr
e
vious
stu
dies
an
d
highli
gh
ts
the
stre
ngths
and
weakn
esses
of
each.
Sect
io
n
3
prese
nts
the
pro
po
se
d
work
in
al
l
it
s
detai
l
s,
an
d
Sect
io
n
4
disc
us
ses
t
he
resu
lt
s.
In
S
ec
ti
on
5
the concl
usi
on
s ar
e i
nclu
ded
.
2.
RELATE
D
W
ORKS
Jai
n
et
al
.
[
11]
intro
duce
d
a
new
m
et
ho
d
for
hi
ding
dat
a
in
an
im
age
,
base
d
on
the
dark
area
s
detect
ed
by
th
e
ed
ge
detect
or
in
t
he
im
age,
an
d
i
nclu
ding
e
ncode
d
te
xt
in
the
le
ast
sign
ific
a
nt
bit.
Th
e
stren
gth
of
this
appro
ac
h
co
m
es
fr
om
us
ing
gr
ay
scal
e
wi
th
edg
e
detect
i
on
a
nd
em
bed
ding
us
in
g
the
LSB
m
et
ho
d
c
om
bi
ned
with
rand
om
e
m
bed
ded
l
eads
to
high
c
onfide
ntial
it
y
a
nd
c
reates
an
e
m
bed
din
g
im
age
just
li
ke
the
or
i
gina
l
i
m
age.
A
ror
a
an
d
An
a
nd
[
12
]
us
e
d
a
s
pa
ti
al
do
m
ai
n
te
chn
i
qu
e
f
or
im
age
ste
ga
nograp
hy
syst
e
m
to
con
ceal
the
te
xt
into
the
col
or
im
a
ges
us
i
ng
t
he
e
dg
e
detect
ion
way,
ed
ges
of
an
im
age
are
detect
ed
by
scan
ning
usi
ng
a
3×
3
wi
ndow,
a
fter
th
at
the
edg
e
pi
xels
was
ra
ndom
izing
by
usi
ng
s
or
ti
ng
m
et
hod,
finall
y,
the
bl
ue
c
om
po
ne
nt
of
s
or
te
d
e
dg
e
pi
xels
e
ncod
ed
t
he
te
xt.
A
nove
l
cha
nne
l
-
de
pende
nt
pa
yl
oad
par
ti
ti
on
strat
e
gy
based
on
a
m
pl
ify
ing
cha
nn
el
m
od
ific
at
ion
pr
ob
a
bili
ties
by
Lia
o
et
al
.
[13]
is
pro
po
s
ed
,
to
adap
ti
ve
assig
n
the
em
bedding
ca
pacit
y
am
ong
R
GB
cha
nnel
s.
A
com
pr
e
ssion
-
ba
sed
ste
gano
gr
a
phy
i
de
a
was
pro
po
se
d
by
Ca
rp
e
ntieri
et
al
.
[
14
]
that
depend
e
d
on
al
gor
it
h
m
s
and
par
a
m
et
ers
us
ed
t
o
create
an
d
m
a
intai
n
the
com
pr
esse
d
ar
chi
ve
e
xp
l
oited,
i
n
a
dd
it
i
on
to
t
he
hiera
rch
ic
al
c
om
pr
essed
a
rc
hiv
e
st
ru
ct
ur
e
it
sel
f,
s
o
that
secret
inf
or
m
at
ion
is
not
se
m
antic
al
ly
rela
te
d
to
the
c
on
t
ents
of
that
co
m
pr
essed
arc
hi
ve.
Alam
et
al
.
[15]
pro
po
se
d
a
ne
w
schem
e
reli
e
s
on
us
in
g
a
secret
key
that
gen
erates
ra
ndom
nu
m
ber
s
us
ing
the
c
ha
otic
log
ist
i
c
m
ap
fo
r ran
do
m
co
m
pen
sat
io
n
LSB de
pendi
ng
on
the e
dge p
ixels i
n
the
co
ver
im
age,
the ed
ge dete
ct
or
c
ann
y
was
a
pp
li
ed
to
get
the
e
dge
i
m
age
from
the
gr
ay
im
age,
af
te
r
that
the
im
age
was
div
ide
d
int
o
a
set
of
blo
c
ks
each
one
of
si
ze
n
pix
el
s
an
d
the
first
bit
ho
l
ds
the
sta
tu
s
of
ot
her
pi
xe
ls
1
or
2
bits
a
re
inclu
ded
when
the
pix
el
is
non
-
ed
ge
if
the
pi
xels
is
the
edg
e
pi
xel,
the
1
-
4
pix
el
s
will
place
and
the
nu
m
ber
of
buil
t
-
in
bi
t
s
is
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Paylo
ad
and q
ua
li
ty
au
gm
e
nt
ation u
sin
g
ste
ganogra
phic
opti
miza
ti
on t
ec
hn
i
qu
e…
(
Maf
az
Al
an
e
zi)
531
ta
ken
ra
ndom
l
y
by
ch
aotic
m
ap.
Ba
nik
an
d
Ba
nd
y
op
a
dhya
y
[16]
s
uggest
ed
a
n
i
nnovat
ive
ste
gano
gr
a
phy
wa
y
wh
ic
h
is
a
preprocesse
d
co
ver
im
age
by
edg
e
detect
io
n
al
gorithm
th
at
us
ed
sobel
op
e
rato
r
as
an
edg
e
detect
ion
m
ask
,
the
n
a
sec
ret
m
essage
is
em
bedde
d
with
a
te
chn
iq
ue
of
m
u
lt
iple
bits’
m
od
i
ficat
ion
whic
h
is
base
d
on
cl
assic
LSB
m
od
ific
at
ion
wh
e
re
t
he
le
ast
1
st
bit
2
nd
bit,
3
rd
bit
4
th
bit
an
d
5
th
bit
o
f
ed
ge
ha
ve
bee
n
change
d.
Pa
ra
h
et
al
.
[17]
ad
op
te
d
a
ne
w
m
et
hod
to
hid
e
t
he
data
i
n
the
colo
r
im
ages
by
div
idin
g
th
e
i
m
age
into
th
ree
re
d,
gr
ee
n,
a
nd
bl
ue
pla
nts,
a
nd
us
in
g
a
c
om
po
sit
e
ed
ge
det
ect
or
re
present
ed
by
a
pr
e
witt
,
an
d
cann
y
detect
or
to
cat
egorize
the
i
m
age
pixe
ls
into
ed
ge
pix
el
s
an
d
non
-
ed
ge
pi
xels,
a
nd
t
he
m
e
tho
d
us
e
d
gr
ee
n
a
nd
blu
e
plants
t
o
i
nclu
de
sec
ret
data
wh
il
e
le
avi
ng
the
r
ed
pla
ne
to
us
e
a
s
a
n
in
di
cat
ion
of
the
st
at
e
of
pix
el
s
w
hethe
r
it
is
an
edg
e
or
not,
the
te
xt
has
been
di
vi
ded
into
four
blo
c
ks
an
d
en
cod
i
ng
data
with
the
rivest
ci
phe
r
4
(
RC4
)
al
gorith
m
to
increa
se
c
onfide
ntial
it
y
a
nd
m
ulti
ple
bits
are
i
nclu
de
d
i
n
e
dg
e
pi
xels
wh
il
e
within
on
e
bit
in
a
pix
el
that
do
e
s
not
bel
ong
to
the
e
dge,
t
he
m
et
ho
d
ha
s
pro
ven
it
s
ef
fici
ency
by
com
par
in
g
it
s
exp
e
rim
ent
al
r
esults
with
oth
e
r
stu
dies.
Kaur
et
al
.
[
6]
pro
po
se
d
a
hy
bri
d
a
ppr
oac
h
w
hich
is
a
c
om
bi
natio
n
of
dif
fer
e
nt
te
c
hn
i
qu
e
s
s
uch
a
s
RSA
f
or
secr
et
m
essage
e
nc
ryptio
n,
ca
nny
for
e
dg
e
detect
ion
,
m
at
ching
, an
d
4
LSB
rep
la
cem
ent
for
the
em
bedde
d
proce
s
s,
in
w
hich
a
t
ext
m
essage
was
hidden
i
ns
id
e
in
al
l
la
ye
rs
of
RGB
colo
r
fr
am
es
of
vid
e
o,
to
acc
om
plish
hig
h
-
c
apacit
y
data
and
high
-
qual
it
y
of
ste
go
vi
deo
base
d
on
the
qual
it
y
m
et
rics
PSN
R,
MSE
and
b
it
error
rate
(
BE
R
)
.
Va
nm
at
hi
and
P
rabu
[
18
]
intro
duce
d
a
te
chnolo
gy
that
hid
es
inf
o
rm
at
ion
th
at
us
es
f
uzzy
edg
e
detect
io
n,
cha
otic
e
ncr
y
ption,
a
nd
a
le
ss
im
po
rtant
bi
t
-
ov
e
r
m
et
ho
d.
T
his
m
et
ho
d
al
lows
the
i
nclusi
on
of
m
or
e
co
nf
i
de
ntial
data
th
at
pro
vid
es
m
or
e
secu
rity
,
gi
ving
th
e
syst
em
a
cl
ear
increase
in
t
he
pro
portion
of
PSN
R
from
5%
t
o
9%
m
ax.
Gaura
v
an
d
G
han
e
ka
r
[
19
]
intr
oduce
d
an
a
lgorit
hm
to
hi
de
in
f
or
m
at
ion
us
in
g
t
he
c
an
ny
m
et
ho
d
for
de
fining
t
he
e
dg
e
s
a
nd
m
or
phologica
l
processes
to
optim
iz
e
i
m
ages
based
on
MSB
a
nd
usi
ng
t
he
X
OR
exclusi
ve
sepa
rati
on
te
c
hn
i
que
and
acc
om
plishe
d
that
the
a
ver
a
ge
PSN
R
is
cl
os
e
to
44
a
nd
tha
t
SSI
M
is
0.
998
at
an
est
a
blished
s
pace
of
1.
25
BPP
.
Se
ti
adi
[8
]
su
gge
ste
d
a
m
et
ho
d
that
f
oc
us
es
on
inc
re
asi
ng
the
payl
oa
d
by
inclu
di
ng
the
te
xt
in
th
e
areas
of
the
e
xten
ded
e
dge,
and
by
encodin
g
t
he
da
ta
by
X
OR
opera
ti
ons
on
th
e
te
xt
with
M
SB
an
d
usi
ng
t
he
LSB
m
et
hod
in
em
beddin
g,
t
he
beg
i
nn
i
ng
is
to
inclu
de
t
he
te
xt
in
t
he
e
dge
pix
el
s
as
a
higher
pri
ori
ty
an
d
if
the
te
xt
ha
s
a
re
st
it
is
inc
lud
e
d
in
the
sm
oo
th
ar
eas
The
stre
ngth
of
this
pa
per
c
om
es
fr
om
the
edg
e
sc
al
ing
proce
ss,
wh
ic
h
increa
sed
the
payl
oad b
y
18.
65% whil
e m
ain
ta
inin
g
th
e ste
go im
age q
uali
ty
.
3.
THE
SUGGE
STE
D
W
ORK
In
t
his
pa
pe
r,
e
dg
e
detect
io
n
processes
we
re
app
li
e
d
to
hide
a
la
rg
e
am
ount
of
te
xt
in
t
he
i
m
age
by
the
em
bed
ding
process,
a
nd
we
sug
gested
us
in
g
“Le
na”,
“Fr
uits”,
“C
at
”
and
“
Sail
s”
i
m
ages
as
the
cov
e
r
i
m
age
in
the
em
bed
din
g
pro
cess,
as
they
a
re
sta
ndar
d
te
s
t
i
m
ages
it
is
widely
us
e
d
in
the
fiel
d
of
im
age
processi
ng.
Va
rio
us
e
dg
e
det
ect
or
s
ha
ve
be
en
us
e
d
s
uch
a
s
sobel
,
prewit
t,
r
ob
e
rts,
l
og,
f
uzzy
lo
gic,
c
ann
y,
hybri
d,
dilat
e
hybri
d
(
5
×
5)
a
nd
dilat
e
hybri
d
ed
ge
detect
or
(
10
×
10)
.
Ste
gano
gr
a
phy
is
app
li
ed
in
bo
t
h
gray
and
c
olo
r
im
ages,
to
com
par
e
res
ults
a
nd
dem
on
strat
e
the
m
axim
um
qu
al
it
y
and
payl
oad
obta
ined
.
The flo
wch
a
rt
in Figu
re
1
re
prese
nts
a
gen
e
r
al
o
utli
ne
of
how
the
stega
nograp
hy
works.
Figure
1.
The
fl
ow
c
har
t
of a
ge
ner
al
outl
ine
of how t
he
st
eg
anog
raphy
wor
ks
3.1.
Ed
ge det
ecto
rs
Ed
ges
are
c
omm
on
ly
te
r
m
ed
as
local
featu
re
s
since
they
ca
rr
ie
d
a
l
ot
of
inf
or
m
at
ion
ab
ou
t
diff
e
re
nt
par
ts
in
the
im
age
an
d
detect
ed
from
the
su
dd
e
n
cha
nge
in
the
gr
ay
le
ve
l,
m
or
e
ever
it
reg
ar
ded
the
bor
der
betwee
n
t
he
diff
e
ren
t
pa
rts
in
the
im
age,
be
cause
it
se
pa
ra
te
s
betwee
n
tw
o
disti
nctly
dif
fer
e
nt
par
ts,
w
her
ea
s
edg
e
detect
ors
com
pu
te
the
gradient
m
agn
it
ud
e
f
ollow
i
ng
t
o
a
f
or
m
ula
that
diff
e
rs
f
ro
m
the
detect
or
to
oth
e
r
,
so
m
e
of
them
base
d
on
a
sin
gle
de
rivati
ve
li
ke
r
ob
e
rt,
s
obel
,
pr
e
witt
,
a
nd
can
ny,
w
h
il
e
oth
e
r
was
use
d
the
seco
nd
de
rivati
ve
as
la
placi
an
of
ga
us
sia
n
(L
OG),
if
the
m
agn
it
ude
of
the
gr
a
dient
is
higher
t
han
a
th
re
sh
ol
d
then
the
e
dge
i
s
existe
nce
[20
]
.
Fu
zzy
lo
gic
dep
e
nds
on
t
he
sub
j
ugat
io
n
of
an
im
age
window
of
pix
el
s
to
set
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
5
2
9
-
5
3
9
532
fu
zzy
te
rm
s
th
a
t
hig
hlig
htin
g
al
l
edg
es
relat
ed
to
an
im
age
,
as
instru
m
ental
fu
zzy
te
r
m
s
to
detect
the
relat
ive
pix
el
s’
value
s
wh
ic
h
can
point
to
ed
ge
pre
sence
[
21]
.
A
nothe
r
ty
pe
of
edg
e
detect
or
w
as
a
hybr
i
d
de
te
ct
or
wh
ic
h
is
a
co
m
bin
at
ion
of
m
ixed
detect
or
s
us
in
g
OR
op
erators
bet
wee
n
them
,
in
Tab
le
1
the
ed
ge
de
te
ct
or
s
us
e
d
in
the s
ug
gested
sch
em
e are list
ed wit
h t
he
num
ber
of
edg
e
p
i
xels
detect
ed
.
Table
1
.
T
he
num
ber
of e
dg
e
pix
el
s
detect
ed
b
y S
obel
, R
oberts, Pre
witt
, LO
G
, Can
ny,
Fuzzy
Logic,
Hybr
i
d
and D
il
at
e H
y
bri
d
for gray im
ages: ‘
Lena
’,
‘
Fr
uits
’,
‘Cat
’
a
nd ‘
S
ai
ls’
Edg
e
Detectio
n
Co
v
er
I
m
ag
es
So
b
el E
d
g
e
Detector
8055
6126
1
0
8
8
2
1
1
6
5
8
Ro
b
erts E
d
g
e
Detector
7785
6000
7915
8940
Prewitt
Edg
e
Detector
7995
6034
1
0
9
0
7
1
1
5
8
6
LOG
Edg
e
Detector
8984
8210
1
4
2
0
2
1
2
0
8
2
Fu
zzy
Log
ic
Edg
e
Detector
1
1
3
7
8
2541
2
1
9
2
5
1
6
8
4
6
Can
n
y
E
d
g
e
Detector
2
2
4
3
8
2
4
3
1
9
3
5
0
0
0
4
1
8
1
7
Hy
b
rid Edg
e
Detector
3
9
2
8
1
3
2
6
1
1
6
1
4
9
9
6
1
0
1
1
Dilate H
y
b
rid
Edg
e
Detector
(5×5
)
9
8
9
7
3
8
5
4
3
8
1
1
9
4
0
8
1
4
5
1
5
2
Dilate H
y
b
rid
Edg
e
Detector
(10
×1
0
)
1
2
7
3
8
1
1
1
7
9
7
6
1
4
9
4
4
5
1
8
0
2
1
3
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Paylo
ad
and q
ua
li
ty
au
gm
e
nt
ation u
sin
g
ste
ganogra
phic
opti
miza
ti
on t
ec
hn
i
qu
e…
(
Maf
az
Al
an
e
zi)
533
3.2.
The
St
e
gara
phic me
thod in
g
r
ay
First,
the
ori
gi
nal
co
ver
im
a
ge
is
div
ide
d
i
nto
64×
64
bloc
ks
,
these
blo
c
ks
are
s
or
te
d
ba
sed
on
the
edg
e
s’
in
f
or
m
a
ti
on
ext
racted
from
app
ly
ing
edg
e
detect
io
n
on
t
he
ori
gi
na
l
cov
e
r
im
age,
the
i
m
age’
s
bl
ock
s
order
is
t
he
fir
st
par
t
of
the
ke
y.
Seco
nd,
w
e
div
i
de
d
t
he
te
xt
into
blo
c
ks
su
it
able
to
the
i
m
age
blo
c
k’
s
siz
e
,
and
the
n
these
blo
c
ks
a
re
s
or
t
ed
acc
ordin
g
t
o
their
data,
he
re
the
te
xt
bl
oc
ks
order
co
ns
i
der
s
the
sec
on
d
par
t
of
the
key.
I
n
a
dd
it
io
n
to
the
key,
the
sel
ect
ed
ed
ge
detect
i
on
ty
pe
is
in
th
e
deal
betwee
n
the
two
par
ti
e
s.
The
te
xt
was
e
m
bed
de
d
in
4
bitp
la
nes
if
edg
e
pix
el
,
an
d
in
2
bitplanes
if
non
-
e
dg
e
pi
xel,
see
the
flow
c
har
t
in
Figure
2.
Figure
2
.
The
fl
ow
c
har
t
of the
suggest
ed
em
beddin
g p
hase
in the g
ray im
a
ge
3.3.
The
s
te
gan
og
r
aphi
c m
eth
od
in
co
l
or i
mag
es
To
hid
e
te
xt
in
a
col
or
e
d
im
a
ge,
first
div
i
de
the
te
xt
int
o
th
ree
par
ts,
each
par
t
of
them
is
treat
ed
as
a
gr
ay
im
age
an
d
eac
h
par
t’
s
si
ze
as
the
c
over
i
m
age’
s
siz
e,
wh
ic
h
is
th
e
fir
st
par
t
of
the
ke
y,
al
so
se
pa
ra
te
the
cov
e
r
im
age
in
to
three
im
ages
Y,
Cb
,
an
d
Cr,
so
t
hat
each
par
t
of
the
te
xt
will
be
e
m
bedded
i
n
one
of
these
three im
ages. The
process
of
e
m
bed
di
ng is a
s foll
ow
s
fo
r
th
e i
m
age and
te
xt p
a
rt:
Determ
ine the
edg
e
d
et
ect
io
n t
ype b
et
wee
n
t
he
tw
o pa
rtie
s.
Divid
e
the
ori
gin
al
c
over
im
age
i
nto
64×6
4
blo
c
ks
,
s
or
t
these
blo
c
ks
ba
sed
on
t
he
e
dges’
in
form
at
i
on
extracte
d
f
ro
m
app
ly
in
g
e
dge
detect
ion
on
t
he
ori
gin
al
co
ve
r
im
age,
the
im
age
blo
c
ks
ord
er
is
t
he
sec
on
d
par
t
of the
key.
Divid
e
the
te
xt
into
blo
c
ks
s
ui
ta
ble
to
the
i
m
age
blo
c
k’s
siz
e,
sort
these
blo
c
ks
acc
ord
ing
t
o
their
dat
a,
the text
blo
c
ks’
order co
ns
ide
rs
the
thir
d part
of the
k
ey
.
Af
te
r
c
om
pletin
g
the
em
beddin
g
process
f
or
the
thr
ee
im
ages,
m
erg
e
them
into
on
e
i
m
age
to
get
back the c
ol
or
e
d
im
age,
w
hich
r
e
pr
ese
nts the
ste
go im
age,
see the
flo
wch
a
rt
in
Fi
gure
3.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
5
2
9
-
5
3
9
534
Figure
3
.
The
fl
ow
c
har
t
of the
suggest
ed
em
beddin
g p
hase
in the c
olor im
age
4.
E
X
PERI
MEN
TATION
R
E
S
ULTS
AND D
ISCUSS
IO
N
The
e
xperim
e
ntati
on
res
ults
are
e
xecu
te
d
to
assess
the
ap
proach
pe
r
form
ance
by
us
in
g
s
om
e
sta
nd
a
rd
te
sts:
The
em
bed
di
ng
capaci
ty
(
pa
yl
oad
)
is
m
easur
e
d
by
the
m
axim
u
m
nu
m
b
er
of
em
bed
di
ng
bits
per pi
xel (
BP
P
) define
d
a
s in
(
1
)
[22]
:
=
∗
(1)
Wh
e
re
H
a
nd
W
a
re th
e
origi
nal cove
r
im
age h
ei
ght a
nd w
i
dth
res
pecti
vel
y.
Peak
sig
nal
-
to
-
noise
rati
o
(PSNR)
m
easur
em
ent
to
cal
cul
at
e
the
qu
al
it
y
dif
fer
e
nce
be
tween
t
he
or
i
gin
al
i
m
ages
and
the
ste
go
i
m
ages,
hig
he
r
PSN
R
value
m
eans
le
ss
distor
ti
on.
I
f
PS
N
R
is
m
or
e
than
4
0
dB
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Paylo
ad
and q
ua
li
ty
au
gm
e
nt
ation u
sin
g
ste
ganogra
phic
opti
miza
ti
on t
ec
hn
i
qu
e…
(
Maf
az
Al
an
e
zi)
535
(d
eci
bels),
the
n
it
is
ve
ry
go
od,
if
PS
NR
be
tween
30
dB
to
40
dB,
t
hen
it
is
acce
ptable
,
an
d
if
it
le
ss
than
30
dB the
n
it
is
not acc
eptable
be
cause t
he dist
ort
ion w
ould
be hig
h [23]
,
[24].
PSNR de
fine
d as i
n
(
2
)
[
25
]
:
PSNR
=
10
×
log
10
255
×
25
5
(2)
Wh
e
re th
e MS
E is the
m
ean sq
ua
re error b
et
ween
th
e p
ixel
s in
the o
ri
gin
a
l im
age an
d
ste
go
im
age,
def
i
ned
as
in
(
3
)
[
26]
:
=
∑
∑
(
−
′
)
2
(
∗
)
=
1
=
1
(3)
Wh
e
re
H
a
nd
W
are
the
ori
gi
nal
cov
e
r
im
a
ge
or
ste
go
im
age
hei
gh
t
a
nd
width
res
pecti
vely
,
an
d
′
refe
r
to
pix
el
values
of
the
ori
gin
al
and
t
he
ste
go
i
m
ages,
res
pec
ti
vely
.
Obviou
sly
,
a
hig
he
r
P
SN
R
m
eans
a
bette
r
qu
al
it
y
that
the
ste
go
im
age
is
ve
ry
cl
os
e
t
o
t
he
or
i
gin
al
im
a
ge
[26].
If
t
he
ste
go
i
m
age
beco
m
es
near
er
to
th
e
cov
e
r
im
age,
the
value o
f
MS
E m
ini
m
iz
es and PSNR m
axim
iz
es [
27
]
,
[28]
.
In
case
th
e
s
ugge
ste
d
sc
hem
e
is
exec
uted
us
in
g
values
of
non
E
dgeB
it
Plane
=
2
a
nd
Ed
geBit
Plane
=4,
e
dge
detect
ion
(
f
uzzy
lo
gi
c,
can
ny,
hybri
d
dilat
e
(
5×5)
an
d
dilat
e
(
10
×10))
on
four
gr
ay
im
ages:
‘Len
a’
,
‘F
r
uits’,
‘Cat
’
and
‘
Sail
s’
of
siz
e
512×51
2for
co
nd
uct
the
exp
e
rim
ents.
The
goal
is
to
em
bed
2
12
B
and
2
13
B,
al
l
po
ssible
ca
ses
of
the
s
uggested
sc
hem
e
are
li
ste
d
in
Table
2.
I
n
th
e
case
te
xt
Size
(
2
12
=
40
064B
)
,
th
e
su
ggest
e
d
sc
he
m
e
achieves
hi
ding
al
l
te
xt
da
ta
wh
ic
h
is
t
he
m
axi
m
u
m
Pa
yl
oad
1.223
bpp
on
al
l
i
m
ages
with
al
l
edg
e
detect
or
s
,
in
the
sam
e
tim
e
it
m
ai
nt
ai
ns
the
im
age
qu
al
it
y,
w
her
e
,
PS
NR
is
hi
gh
f
ro
m
43
.
003dB
to
45.79
0
dB
by
us
in
g
a
f
uzzy
log
ic
detect
or.
In
ot
her
cases
,
te
xt
Size
(
2
13
=
8012
8B)
t
he
m
axi
m
u
m
p
os
sible
payl
oad
va
lue
is
2.4
45bpp,
a
nd
the
res
ulti
ng
payl
oad
f
or
i
m
ages
ra
nge
f
r
om
2.
019b
pp
t
o
2.445
bpp
de
pend
on
the
edg
e
detec
tor
us
e
d
an
d
how
m
any
the
i
m
age
con
ta
in
ed
ges.
T
he
m
axi
m
u
m
pay
load
(
2.4
45bp
p)
was
achieve
d
by
usi
ng
dilat
e
edge
detect
or
(
5×
5)
a
nd
(
10×1
0)
with
the
‘
Sail
s’
i
m
age
wh
il
e
the
i
m
age
qu
al
it
y
rem
ai
ns
acceptable wit
h
PS
N
R
= 35.
286d
b.
To
see
the
difference
bet
wee
n
em
beddin
g
m
or
e
in
the
edg
e
or
non
-
e
dg
e
bette
r,
her
e
the
sugg
e
ste
d
schem
e
is
execu
te
d
us
in
g
valu
es
of
non
E
dge
Bi
tPl
ane
=
4
a
nd
Ed
geBit
Plane
=
2,
ed
ge
de
te
ct
ion
(fuzzy
log
ic
,
cann
y,
hybri
d
dilat
e
(5
×
5)
a
nd
dilat
e
(
10
×
10)
)
on
f
our
gr
a
y
i
m
ages:
‘Lena’,
‘
Fruit
s’,
‘Cat
’
an
d
‘S
ai
ls’
of
siz
e
512×
512
for
c
onduct
the
e
xperim
ents.
As
sh
ow
n
in
Tabl
e
3,
al
t
hough
t
he
s
ugge
ste
d
s
chem
e
achieves
th
e
sam
e
m
axi
m
u
m
p
ay
load
of 1.2
23
bpp wit
h
a text si
ze of
(400
64
B
),
in
c
ontrast
, th
e P
SNR
v
al
ues
dec
re
ase, f
or
exam
ple, the
P
SN
R val
ue dec
rease fro
m
4
4.401 (Table
2)
t
o
37.
023d
b
(T
able 3) for
t
he ‘Lina
’
im
age
with the
fu
zzy
l
og
ic
det
ec
tor
a
nd all
o
t
her im
ages.
Finall
y,
ap
plyi
ng
the
s
uggest
ed
sc
hem
e
on
colo
r
im
ages,
wh
e
re,
it
is
e
xe
cuted
us
i
ng
va
lues
of
non
Ed
geBit
Plane
=
2
a
nd
E
dg
e
Bi
tPl
ane
=
4,
edg
e
detect
io
n
(fuzzy
lo
gic,
cann
y,
hy
br
i
d
dilat
e
(5
×
5))
on
f
our
-
colo
r
im
age
s: ‘Len
a’
, ‘Fr
uit
s’, ‘Cat
’
a
nd ‘Sai
ls’
of
siz
e 51
2×
512 f
or
c
onduc
ti
ng
the
expe
rim
ents.
As
s
how
n
in
Table
4,
in
t
he
case
te
xt
siz
e
(
212
=
4006
4B),
the
s
ugge
ste
d
sc
hem
e
achieves
te
xt
hid
in
g
with
m
axim
u
m
payl
o
ad
1.2
23bp
p
on
al
l
i
m
ages
with
al
l
edg
e
det
ect
or
s,
i
n
the
s
a
m
e
tim
e
i
t
m
a
intai
ns
the
im
age
qu
al
it
y,
wh
er
e,
P
S
NR
is
higher
f
ro
m
40
.
579
dB
to
41.76
6
dB
by
us
i
ng
the
c
ann
y
detect
or.
So
,
we
increase
t
he
hi
dd
e
n
te
xt
siz
e
un
ti
l
(
199680
B)
w
her
e
t
he
m
axi
m
u
m
paylo
ad
f
or
it
is
6.096
BP
P
,
the
resu
lt
s
sh
ow
that
al
m
os
t
al
l
i
m
ages
reac
h
the
m
axim
u
m
payl
oad
with
acce
pta
ble
PSN
R
in
thi
rtie
s
values.
T
able
5
sh
ows
there
is
no appa
ren
t
dis
tortio
n
in
the i
m
ages af
te
r
h
i
ding
payl
oad
m
or
e than
6
B
PP
.
Table
2
.
T
he
e
xp
e
rim
entat
ion
outcom
es o
f
t
he
s
uggeste
d
s
chem
e u
ti
li
zi
n
g values
of
Non
-
E
dgeB
it
Plan
e
=
2
and E
dg
eB
it
Pl
ane =
4 o
n gr
a
y im
ages ‘
Le
na
’,
‘Fruits’
, ‘
C
at
’
an
d ‘Sail
s’,
gr
ay
im
ages s
iz
e of
512×
512
Edg
e Dete
ctio
n
Text Size
(
B
y
te)
Lena
Fruits
Cat
Sails
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
Fu
zzy
Log
ic
4
0
0
6
4
B
4
4
.40
1
1
.22
3
4
5
.79
0
1
.22
3
4
3
.00
3
1
.22
3
4
3
.64
9
1
.22
3
Can
n
y
4
0
0
6
4
B
4
2
.96
2
1
.22
3
4
2
.87
6
1
.22
3
4
2
.04
1
1
.22
3
4
1
.40
6
1
.22
3
Hy
b
rid
4
0
0
6
4
B
4
1
.78
2
1
.22
3
4
2
.17
2
1
.22
3
4
0
.74
9
1
.22
3
4
0
.49
4
1
.22
3
Dilate (
5
×5
)
4
0
0
6
4
B
3
9
.47
8
1
.22
3
3
9
.80
5
1
.22
3
3
9
.19
8
1
.22
3
3
8
.27
4
1
.22
3
Dilate(10
×1
0
)
4
0
0
6
4
B
3
8
.81
5
1
.22
3
3
8
.92
8
1
.22
3
3
8
.75
5
1
.22
3
3
7
.81
1
1
.22
3
Fu
zzy
Log
ic
8
0
1
2
8
B
4
2
.08
6
2
.08
7
4
3
.58
0
2
.01
9
4
0
.42
1
2
.16
6
4
1
.13
5
2
.12
9
Can
n
y
8
0
1
2
8
B
4
0
.35
6
2
.17
0
4
0
.26
6
2
.18
6
3
9
.22
0
2
.25
6
3
8
.66
0
2
.31
9
Hy
b
rid
8
0
1
2
8
B
3
9
.07
8
2
.26
2
3
9
.46
7
2
.24
5
3
7
.77
0
2
.31
2
3
7
.56
8
2
.39
9
Dilate (
5
×5
)
8
0
1
2
8
B
3
6
.55
2
2
.38
9
3
6
.78
5
2
.39
6
3
6
.10
5
2
.35
7
3
5
.28
6
2
.44
5
Dilate(10
×1
0
)
8
0
1
2
8
B
3
5
.85
1
2
.41
7
3
5
.94
8
2
.41
4
3
5
.56
1
2
.37
0
3
4
.78
9
2
.44
5
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
5
2
9
-
5
3
9
536
Table
3
.
T
he
e
xp
e
rim
entat
ion
outcom
es o
f
t
he
s
uggeste
d
s
chem
e u
ti
li
zi
n
g values
of
NonE
dg
eB
it
Plane
=4
a
nd
Ed
geBit
Plane
=2 on g
ray im
a
ges ‘Le
na’
,
‘Fr
uits’, ‘Ca
t’ a
nd
‘
Sail
s’
, gray i
m
ages s
iz
e of
512×
512
Edg
e Dete
ctio
n
Text Size
(By
te
)
Lena
Fruits
Cat
Sails
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
Fu
zzy
Log
ic
4
0
0
6
4
B
3
7
.02
3
1
.22
3
3
6
.99
1
1
.22
3
3
7
.21
9
1
.22
3
3
7
.11
5
1
.22
3
Can
n
y
4
0
0
6
4
B
3
7
.17
4
1
.22
3
3
7
.19
6
1
.22
3
3
7
.40
0
1
.22
3
3
7
.32
9
1
.22
3
Hy
b
rid
4
0
0
6
4
B
3
7
.24
8
1
.22
3
3
7
.27
1
1
.22
3
3
7
.69
4
1
.22
3
3
7
.54
2
1
.22
3
Dilate (
5
×5
)
4
0
0
6
4
B
3
8
.05
9
1
.22
3
3
7
.85
7
1
.22
3
3
8
.66
0
1
.22
3
3
8
.83
5
1
.22
3
Dilate(10
×1
0
)
4
0
0
6
4
B
3
8
.62
0
1
.22
3
3
8
.39
5
1
.22
3
3
9
.41
4
1
.22
3
3
9
.85
9
1
.22
3
Table
4
.
T
he
e
xp
e
rim
entat
ion
outcom
es o
f
t
he
s
uggeste
d
s
chem
e u
ti
li
zi
n
g values
of
Non
-
E
dgeB
it
Plan
e=4
and E
dg
eB
it
Pl
ane =
2 on col
or im
ages ‘
Len
a
’,
‘F
r
uits’
, ‘
Ca
t
’
a
nd ‘
Sail
s’
, c
olor im
ages s
iz
e of
512×
512
Edg
e Dete
ctio
n
Text Size
(
B
y
te)
Lena
Fruits
Cat
Sails
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
PSN
R
(dB
)
Pay
lo
ad
(bp
p
)
Fu
zzy
Log
ic
4
0
0
6
4
B
3
9
.98
2
1
.22
3
4
0
.36
5
1
.22
3
4
1
.59
3
1
.22
3
3
9
.81
5
1
.22
3
Can
n
y
4
0
0
6
4
B
4
1
.76
6
1
.22
3
4
1
.59
5
1
.22
3
4
0
.57
9
1
.22
3
4
0
.92
5
1
.22
3
Hy
b
rid
4
0
0
6
4
B
3
9
.21
2
1
.22
3
3
9
.38
1
1
.22
3
3
9
.56
7
1
.22
3
3
8
.77
6
1
.22
3
Dilate (
5
×5
)
4
0
0
6
4
B
3
9
.06
9
1
.22
3
3
9
.17
6
1
.22
3
3
8
.43
3
1
.22
3
3
8
.25
9
1
.22
3
Fu
zzy
Log
ic
1
9
9
6
8
0
B
3
3
.20
1
6
.07
8
3
3
.63
9
6
.06
5
3
4
.90
0
6
.07
4
3
3
.05
6
6
.07
2
Can
n
y
1
9
9
6
8
0
B
3
5
.04
5
6
.09
0
3
4
.82
7
6
.09
0
3
3
.81
4
6
.08
9
3
4
.19
9
6
.08
9
Hy
b
rid
1
9
9
6
8
0
B
3
2
.37
6
6
.09
2
3
2
.57
0
6
.09
1
3
2
.73
7
6
.08
9
3
1
.94
7
6
.09
6
Dilate (
5
×5
)
1
9
9
6
8
0
B
3
2
.10
4
6
.09
2
3
2
.15
0
6
.09
1
3
1
.53
8
6
.08
9
3
1
.35
7
6
.09
0
Table
5.
Sho
w t
he
ex
pe
rim
ent
at
ion
ou
tc
om
es
of the
sug
gest
ed
sc
hem
e w
it
h
se
ver
al
e
dge
detect
or
s
Origin
al
I
m
ag
e
Fu
zzy
Can
n
y
Hy
b
rid
Dilate
PSN
R
=
Pay
lo
ad
=
3
2
.13
9
d
B
6
.08
8
b
p
p
3
3
.06
9
d
B
6
.09
4
b
p
p
3
1
.05
7
d
B
6
.09
4
b
p
p
3
0
.40
7
d
B
6
.09
4
b
p
p
PSN
R
=
Pay
lo
ad
=
3
3
.05
6
d
B
6
.07
2
b
p
p
3
4
.19
9
d
B
6
.08
9
b
p
p
3
1
.94
7
d
B
6
.09
1
b
p
p
3
1
.35
7
Db
6
.09
0
b
p
p
Figure
4
s
how
s
a
com
par
ison
of
the
outc
om
e
s
PSN
R
of
gr
a
y
i
m
ages,
'
Lena'
,
'
Fr
uits'
,
'
Ca
t
'
and
'
Sail
s'
,
i
m
ages
siz
e
of
512
×
512
uti
li
zi
ng
values
of
Non
-
E
dg
eB
it
Plane
=
2
an
d
Ed
geBit
Plane
=
4,
em
bed
ded
te
xt
siz
e
=
"400
64B"
.
Also
a
c
om
par
ed
the
PS
NR
res
ults
f
or
the
“4
0064B”
or
“
8012
8B”
e
m
bed
ded
te
xt
siz
e.
I
n
the
'
51
2
×
512'
'
Lena'
gr
ay
i
m
age
usi
ng
N
on
-
Ed
geBit
Plane
=
2
an
d
E
dgeB
it
Plane
=
4,
as
sh
ow
n
in
Fi
gure
5.
The
com
par
is
on
was
m
ade
between
the
two
m
et
ho
ds
of
em
bed
di
ng,
the
first
m
eth
od
in
w
hich
te
xt
i
s
e
m
bed
de
d
in
4
bitplane
s
if
it
is
pix
el
edg
e,
and
in
2
bits
if
the
pix
el
is
no
n
-
e
dg
e
,
and
t
he
seco
nd
m
eth
od
of
e
m
bed
di
ng
i
n
2
bitpla
nes
if
it
is
an
edg
e
pi
xel,
an
d
in
4
bi
tplanes
if
it
is
a
non
-
e
dge
pi
xel,
by
c
om
par
ing
the
PSN
R re
su
lt
s ag
ai
ns
t a 5
12
×
512
Le
na
gray
i
m
age,
e
m
bed
d
ed
te
xt size = “40
064B” see t
he
grap
h
in Figure 6
and
Fi
gure
7
it
is
a
co
m
par
ison
of
the
outc
om
es
PSN
R
on
'
Lena'
gr
ay
i
m
age
and
'
Len
a'
color
i
m
age,
siz
e
of
512
×
51
2 uti
lizing
value
s
of
Non
-
Ed
geBit
P
la
ne
=
2
a
nd E
dg
eB
it
Plane
=
4,
em
bedded te
xt size = “
4006
4
B”
.”
To
e
valuate
th
e
perform
ance
of
t
he
pr
opos
e
d
m
et
ho
d,
a
c
om
par
at
ive
analy
sis
was
pe
rfo
rm
ed
base
d
on
P
SN
R
,
w
hich
m
easur
e
the
per
ce
ntage
di
stortions
in
pe
rcep
ti
on,
to
m
easur
e
the
pe
rfor
m
ance
of
var
i
ous
m
et
ho
ds
by
usi
ng
un
i
form
exp
erim
ental
set
t
ing
s
,
base
d
on
so
m
e
sa
m
e
par
a
m
et
er
li
ke
ste
go
im
age,
te
ch
nical
pro
per
ti
es,
an
d
secur
it
y
aspe
ct
s.
Fo
r
e
xp
e
ri
m
entat
ion
s
both
colo
r
an
d
gr
ey
scal
e
Lena.png
a
nd
Ba
bo
on.
png
wer
e
us
ed
as
cov
e
r
im
ages
,
with
dim
ension
s
of
512×
512
pi
xels
f
or
the
two
im
a
ges.
Ta
ble
6
sh
ow
s
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Paylo
ad
and q
ua
li
ty
au
gm
e
nt
ation u
sin
g
ste
ganogra
phic
opti
miza
ti
on t
ec
hn
i
qu
e…
(
Maf
az
Al
an
e
zi)
537
com
pa
risons
of
PS
NR
value
s
for
the
m
et
ho
ds
r
oj
al
i
the
schem
e
wh
ic
h
us
ed
a
m
od
if
ic
at
ion
VIGE
NERE
ci
ph
e
r,
LSB
m
et
hod,
an
d
dict
ion
a
ry
-
ba
sed
c
om
pr
essio
n
m
e
thod,
Y
U
NG
t
he
schem
e
us
ed
k
-
m
eans
al
go
rithm
for
‘traini
ng
the
palet
te
’,
EZ
-
ste
go
sche
m
e,
Fr
id
rich
the
sc
hem
e
us
ed
palet
te
-
ba
sed
ste
gano
graph
y
m
et
ho
d [
10
]
, a
nd pr
opos
e
d
sc
hem
e
.
Figure
4. Com
par
is
on of t
he ou
tc
om
es PSNR
on 4
gr
ay
im
ages
Figure
5. Com
par
i
ng PSNR
r
esults ba
sed
on
the
e
m
bed
de
d
te
xt
siz
e
Figure
6. Com
par
is
on of
PSNR
r
es
ults acc
or
ding
to the em
beddi
ng m
et
ho
d
Figure
7. Com
par
e
PSNR
res
ults base
d o
n g
ray o
r
c
olor
i
m
ages
Table
6.
T
he
c
om
par
isons
of
PSN
R
val
ues f
or sev
e
ral
wor
ks
with t
he p
r
opose
d m
et
ho
d
Co
v
er
I
m
ag
e
Ro
jali Sche
m
e
YUNG
Sch
e
m
e
EZ
-
steg
o
Sch
e
m
e
Fridrich
Sche
m
e
Prop
o
sed
Metho
d
Lena
51
37
14
31
51
Bab
o
o
n
51
36
15
1
51
5.
CONCL
US
I
O
NS
In
this
pa
per,
a
ste
gan
og
raphy
app
r
oac
h
em
p
loyi
ng
ed
ge
de
te
ct
ion
was
sugg
e
ste
d.
T
his
schem
e
is
a
good
ch
oice
be
cause
it
pr
e
ve
nts
the
hum
an
ey
e
fr
om
no
ti
ci
ng
any
diff
e
re
nce
in
the
ste
go
i
m
age.
O
ur
go
al
is
edg
e
detect
io
n
and
car
ries
a
great
er
nu
m
ber
of
em
bed
ding
bits
to
ac
hieve
high
payl
oa
d
without
a
ff
ect
ing
th
e
i
m
age
cl
arit
y
a
nd
qu
al
it
y
and
i
m
pr
ovin
g
the
eff
ic
ie
ncy
of
th
e
syst
e
m
no
t
only
in
te
r
m
s
of
payl
oad
a
nd
qual
it
y
bu
t al
so
in
te
r
m
s
o
f
secur
it
y and
c
onfide
ntial
it
y
of
the h
id
den
in
f
or
m
at
ion
. A
lt
ho
ugh
st
eg
an
ogra
phy i
m
ages in
the
s
patia
l
dom
ai
n
are
us
ua
ll
y
easy
to
be
at
ta
cked
a
nd
pe
ne
trat
ed,
it
pro
vi
des
m
or
e
payl
oads.
For
this
r
easo
n,
we
ha
ve
fo
c
use
d
ou
r
w
ork
on
inc
reasin
g
c
onfide
ntial
it
y
and
secu
rity
by
pr
e
par
in
g
a
secur
e
c
omm
un
ic
at
ion
te
chn
iq
ue
that
has
be
e
n
pr
e
pa
red
that
c
on
ta
i
ns
a
key
co
ns
i
sti
ng
of
tw
o
or
three
pa
rts
use
d
in
the
proc
ess
of
e
m
bed
di
ng
inf
or
m
at
ion
in
im
ages
gr
ay
a
nd
colo
red
res
pec
ti
vely
.
In
a
ddit
ion
t
o
ag
reein
g
at
the
be
ginni
ng
of
the d
eal
on
the ty
pe
of
e
dge de
te
ct
or
to
b
e u
s
ed
i
n
the
em
bed
di
ng p
r
ocess.
Ex
per
im
ents
w
ere p
er
form
ed
us
in
g
edg
e
pix
el
s
de
te
ct
ed
by
sobe
l,
rober
ts
,
pre
witt
,
LO
G,
ca
nn
y,
fu
z
zy
logi
c,
hybri
d,
dilat
e,
and
hybri
d
on
both
gr
ay
a
nd
c
olor
ed
im
ages.
T
he
la
rg
e
e
dg
e
pi
xels,
res
ulti
ng
from
the
us
e
of
hybr
i
d
a
nd
f
uzzy
detect
ion
al
low
m
or
e
secret
da
ta
to
ac
hieve
higher
payl
oa
ds,
s
o
it
ca
n
be
sai
d
t
hat
the
work
c
on
tri
bu
t
es
to
ge
ner
at
in
g
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
5
2
9
-
5
3
9
538
la
rg
est
possibl
e
num
ber
of
edg
e
s
to
hid
e
the
data.
Th
e
ex
per
im
ental
res
ults
sho
w
that
the
P
SNR
was
increase
d
f
ro
m
a
m
ini
m
u
m
o
f
5%
t
o
a
m
a
xim
u
m
of
8%,
wh
ic
h
m
eans
that
the
sugge
ste
d
ste
ga
nograp
hy
te
chn
iq
ue
a
ppr
oach
is
the
bes
t
in
te
rm
s
of
pa
yl
oad
,
c
onfi
de
ntial
it
y
as
well
as
qu
al
it
y
so
that
the
ste
go
i
m
age
app
ea
rs
as
cl
ose
as
possible
to
the
ori
gin
al
i
m
age.
In
the
f
uture,
ou
r
ap
proach
m
ay
be
app
li
ed
a
fter
m
akin
g
enh
a
ncem
ents
to
im
ages
or
c
anceli
ng
no
ise
from
the
m
,
an
d
ste
gano
gr
a
phy
m
ay
be
ap
pl
ie
d
in
t
he
fr
e
quency
do
m
ai
n
of the
im
age
.
ACKN
OWLE
DGE
MENTS
The
resea
rch
e
r
s
tha
nk
the
D
epar
tm
ent
of
Com
pu
te
r
Sci
ence,
Coll
ege
of
Com
pu
te
r
Scie
nce
a
nd
Ma
them
a
ti
cs, U
ni
ver
sit
y o
f M
os
u
l.
REFERE
NCE
S
[1]
Mare
,
S.
F.,
Vla
duti
u,
M.,
and
Prodan,
L
.
,
“
High
ca
pacit
y
steg
an
ogra
phic
a
lgori
t
hm
base
d
on
pa
y
lo
ad
ada
p
ta
t
ion
and
opti
m
izati
o
n,
”
7th
IEEE
I
nte
rnational
Sy
mpos
ium
on
Appl
ie
d
Computa
tional
Intelli
g
ence
and
Informati
cs
(
SACI
)
,
2021,
pp
.
87
–
92
,
doi
:
10
.
1109/sac
i
.
2012.
6249981.
[2]
D.
Setiadi
and
J.
Jum
ant
o,
“
An
Enha
n
ce
d
LSB
-
Im
age
Stega
n
ogra
ph
y
Us
ing
the
H
y
brid
C
an
n
y
-
Sobel
Edg
e
Dete
c
ti
on,
”
Cy
berne
tics
and
Information
Techno
logi
es
,
vol.
18,
n
o.
2,
pp.
7
4
–
88,
2018,
doi:
htt
ps://
do
i.
org/1
0.
2478/cait
-
2018
-
0029.
[3]
S.
Singh
and
A.
Data
r,
“
Im
prove
d
Hash
Based
Approac
h
for
Sec
ure
Color
Im
age
Stega
nogra
ph
y
using
Cann
y
Ed
ge
Dete
c
ti
on
Metho
d,
”
Inte
rnat
ional
Journal
of
Computer
Sci
en
c
e
an
d
Net
work
Sec
ur
it
y
(
IJCSNS)
,
vol.
15,
no.
7
,
pp.
92
–
98,
2015
.
[4]
Kus
um
a,
E.
J.,
I
ndria
ni
,
O.
R.
,
S
ari
,
C
.
A.
,
and
R
ac
hm
awa
nto
,
E.
H.
,
“
An
imperc
e
pti
ble
LSB
imag
e
hidi
ng
on
edge
reg
ion
using
des
enc
r
y
pti
on
,
”
20
17
Inte
rnational
Confe
renc
e
on
Innov
ative
and
Creati
v
e
Information
Technol
og
y
(
ICITec
h)
,
2017,
vol. 2018
,
pp.
1
–
6
,
doi
:
10
.
1109
/INNO
CIT.
2017.
8319132.
[5]
Essa,
R.
J.,
Abd
ull
ah
,
N.
A.
,
and
Al
-
Dabba
gh,
R.
D.
,
“
Stega
nogr
a
ph
y
Techni
que
u
sing
Gene
tic
Alg
orit
hm
,
”
Iraqi
J.
Sci
.
,
vo
l. 59, no.
3A,
2018,
do
i: 1
0.
24996/i
js
.
2018
.
59.
3a
.
19.
[6]
R.
Kaur,
Pooja,V
arsha
.
,
“
A
h
y
b
rid
appr
oac
h
for
vide
o
stega
nog
rap
h
y
using
edg
e
det
e
ct
ion
and
ide
ntica
l
m
at
c
h
te
chn
ique
s,”
20
16
Inte
rnational
Confe
renc
e
on
Wirel
ess
Com
municat
ions,
Si
gnal
Proce
ss
ing
and
Net
working
(
WiSP
NET)
,
2016,
pp
.
867
-
871
,
doi:
10
.
1109/Wi
SP
NET.
2016.
75
66255.
[7]
S.
Arora
and
S.
Anand,
“
A
New
Approac
h
for
I
m
age
Stega
nogr
aph
y
using
Edge
Dete
c
ti
on
Meth
od,
”
Int
ernati
on
al
Journal
of
innovative
R
ese
arch
i
n
compute
r
and
communic
ati
on
Engi
ne
ering
,
vol
.
1
,
no
.
3
,
pp
.
62
6
–
629,
2013
.
[8]
D.
R.
I.
M.
Set
i
adi
,
“
Pa
y
loa
d
en
hanc
ement
on
least
signifi
c
ant
bi
t
image
stega
no
gra
ph
y
using
ed
ge
area
dilati
on
,
”
Int.
J. Ele
c
tron. T
el
ec
omm
un
.
,
v
ol.
65
,
no
.
2
,
pp
.
287
–
292,
2019
,
doi:
10
.
24425/ij
et
.
2019
.
126312.
[9]
Chen,
W
.
J.,
Ch
an
g,
C.
C
.
,
and
Le
,
T.
H.
N.
,
“
High
pa
y
loa
d
st
eg
anogr
aph
y
m
ec
h
ani
sm
using
h
y
b
rid
edge
d
et
e
ct
or
,
”
Ex
pert
Syste
ms
wit
h
appl
ic
a
ti
on
s
,
vol. 37, no. 4,
pp.
3292
–
3301
,
2010,
doi
:
10
.
10
16/j
.
eswa.
2009.
0
9.
050.
[10]
Ans
ari
,
A.
S.,
Moham
m
adi
,
M.
S
.
,
and
Parve
z,
M.
T.
,
“
A
Co
m
par
at
iv
e
Stud
y
of
Rec
en
t
Stega
nogr
aph
y
T
ec
hn
iques
for
Multi
ple
Im
a
ge
Form
at
s
,
”
Inte
rnational
Journ
al
of
Computer
Net
work
and
Information
Sec
urit
y
,
vol.
11
,
no.
1,
pp.
11
–
25
,
2019
,
doi: 10.
5815
/i
j
c
nis.20
19.
01.
02
.
[11]
Jain,
N
.
,
M
eshra
m
,
S.,
and
Dube
y
,
S.
,
“
Im
age
st
ega
nogr
a
ph
y
using
LSB
and
edge
–
d
et
e
ct
ion
technique
,
”
Inte
rnational
Jo
urnal
of
So
ft
Co
mputing
and
En
gine
ering
(
IJS
CE)
,
vol. 2, no. 3,
pp.
217
–
222
,
20
12.
[12]
S.
Arora
and
S.
Anand,
“
A
Pro
posed
Method
for
Im
age
Stega
nogra
ph
y
using
Edge
Detect
io
n
,
”
Int.
J.
Emerg.
Te
chno
l. Adv.
E
ng.
,
vol. 3, no. 2, pp. 22
–
25,
2013
,
doi
:
10
.
31031/r
m
es.
2018.
03.
00
0569.
[13]
Li
ao
,
X.
,
Yu,
Y
.
,
Li,
B
.
,
Li,
Z
.
,
and
Qin,
Z.
,
“
A
New
Pa
y
lo
ad
Pa
rti
ti
on
Strateg
y
i
n
Color
Im
age
Stega
nogra
ph
y
,
”
IEE
E
Tr
ansacti
ons
on
Circui
ts
and
Syste
ms
f
or
Vi
deo
Te
chn
ology
,
vol.
30,
no.
3,
pp.
685
–
696,
2020,
doi
:
10.
1109/T
CS
VT
.
2019.
2896270
.
[14]
Carpe
nt
ie
ri
,
B.
,
Casti
gli
on
e,
A.,
De
Santi
s,
A.,
Palmieri,
F.,
and
Pizzolante,
R.
,
“
Com
pre
ss
ion
-
base
d
stega
nogra
ph
y
,
”
Concurrency
a
nd
Computati
on:
Pract
ice
and
Ex
peri
enc
e
,
vol
.
32,
no.
8,
p.
e
5322,
2020,
doi
:
htt
ps://
do
i.
org/1
0.
1002/c
p
e.
5322
.
[15]
Alam,
S.,
Kumar
,
V.,
Siddiqui
,
W
.
A.,
and
Ahm
ad,
M.
,
“
K
e
y
dep
ende
n
t
image
stega
nogr
aph
y
using
edg
e
det
e
ct
ion
,
”
Int
ernati
onal
Conf
ere
nce
on
Ad
vanced
Computi
ng
&
C
omm
unic
ati
on
Technol
ogie
s
,
2014,
pp.
85
–
8
8
,
doi:
10
.
1109/AC
CT.
2014.
72
.
[16]
B.
G.
Bani
k
and
S.
K.
Band
y
opa
dh
y
a
y
,
“
an
Im
age
Stega
nogra
ph
y
Method
on
Edge
Dete
c
ti
on
Us
ing
Multi
ple
Ls
b
Modific
a
ti
on
Techni
que
,
”
J
.
Bas
ic
Appl
.
R
es.
In
t.
,
vol
.
9
,
no
.
2
,
pp
.
75
–
80
,
2015
.
[17]
Para
h,
S.
A.,
Sheikh,
J.
A.,
Akhoon,
J.
A.,
Loa
n
,
N.
A.,
and
Bhat
,
G.
M.
,
“
Inform
at
ion
hidi
ng
i
n
edge
s:
A
high
ca
pa
ci
t
y
informati
on
hid
ing
tec
hnique
using
h
y
brid
edge
de
tect
ion,
”
Mul
ti
med
i
a
Tools
and
Ap
pli
cations
,
vol
.
77,
no.
1
,
pp
.
185
–
2
07,
2018
,
doi
:
10
.
1007/s11042
-
01
6
-
4253
-
x.
[18]
C.
Vanm
at
hi
an
d
S.
Prabu,
“
Ima
ge
St
ega
nogra
p
h
y
Us
ing
Fuz
z
y
Logi
c
and
Cha
oti
c
fo
r
L
arg
e
P
a
y
loa
d
and
Hig
h
Im
per
ce
pti
b
il
i
t
y
,
”
Int
ernati
onal
J
ournal
of
Fuzzy
Syste
ms
,
vol
.
20
,
no.
2
,
pp
.
460
–
473,
2018,
doi:
10.
1
007/s40815
-
017
-
0420
-
0.
[19]
K.
Gaura
v
and
U.
Ghane
kar
,
“
I
m
age
stega
nogr
aph
y
base
d
on
Cann
y
edge
detec
t
ion,
d
il
a
ti
on
oper
at
or
and
h
ybrid
codi
ng,
”
Journal
of
Information
Sec
urit
y
and
Appl
i
cat
ions
,
vol
.
41,
pp.
41
-
51,
2018,
doi:
10.
1016/j.ji
sa
.
20
18.
05.
001
.
Evaluation Warning : The document was created with Spire.PDF for Python.