Indonesi
an
Journa
l
of
El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
1
,
Jan
uar
y
201
9
,
pp.
147~
154
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
1
.pp
147
-
154
147
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Developi
ng audio
data hid
ing sche
me using
random
sa
mp
le
bits
with lo
gical
operat
or
s
Moham
med
Hatem
Ali Al
-
Hooti
1
, T
ohar
i Ahm
ad
2
,
Supeno
Djanali
3
1,2,3
Depa
rtment
o
f
Inform
at
i
cs,
In
st
it
ut Te
kno
logi
Sepuluh
Nopem
ber
,
Ea
st Java
,
Indone
sia
1
Facul
t
y
of
Com
pute
r
Sc
ie
nc
e an
d
Inform
at
ion
T
ec
hnolog
y
,
San
a
’a
Univer
si
t
y
,
Sa
na’
a, Ye
m
en
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
ul
6
,
201
8
Re
vised
N
ov
2
,
201
8
Accepte
d
Nov
1
9
, 201
8
Sharp
deve
lopment
progre
ss
of
informati
on
tech
nolog
y
has
aff
e
ct
ed
m
an
y
aspe
ct
s
includin
g
dat
a
sec
ur
ity
.
Thi
s
is
bec
ause
cl
assifi
ed
data
are
oft
e
n
tra
nsferre
d
bet
w
ee
n
s
y
s
te
m
s.
In
thi
s
ca
se
,
da
ta
h
idi
ng
ex
ists
to
p
rote
c
t
such
dat
a
.
Som
e
m
et
h
ods
which
hav
e
be
en
proposed
,
howeve
r,
are
not
y
et
opti
m
a
l
conc
ern
ing
th
e
a
m
ount
of
the
sec
ret
and
th
e
qua
li
t
y
of
th
e
resul
te
d
stego
data
.
In
thi
s
pape
r
,
we
expl
ore
an
aud
i
o
fil
e
as
the
m
ed
ium
to
ca
rr
y
th
e
sec
ret
d
ata
which
has
b
ee
n
ext
ra
ct
ed
int
o
bi
nar
y
.
B
efo
re
the
proc
es
s
beg
ins,
the
cove
r
is
conve
rt
ed
to
bi
nar
y
and
each
sam
ple
’s
bit
s
are
divi
ded
int
o
t
wo
groups,
one
is
used
as
t
he
locati
on
of
t
he
embedde
d
4
bit
s
where
as
the
sec
ond
par
t
loc
a
te
s
the
two
bit
s
tha
t
are
ran
dom
l
y
select
ed
as
the
ke
y
.
The
expe
riment
a
l
result
s
have
val
i
dat
ed
that
the
capac
i
t
y
is
high
a
nd
the
re
is
no
muc
h
impact
on
the
qua
li
t
y
.
Moreove
r,
compare
d
to
the
cur
re
nt
LSB
m
et
hods
the
sec
u
r
i
t
y
is e
xc
ee
ding
l
y
e
nhanc
ed
.
Ke
yw
or
d
s
:
Data hi
ding
Data p
r
otect
ion
Inform
at
ion
secur
it
y
Secret
data
Stegan
ogra
phy
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Moh
am
m
ed
H
at
e
m
A
li
A
l
-
H
oo
ti
,
Dep
a
rtm
ent o
f Info
rm
at
ic
s,
In
sti
tut Te
knol
og
i
Sepulu
h N
op
em
ber
,
Su
r
abaya,
S
ukolil
o,
60111,
E
ast
Jav
a
,
In
done
sia
.
Em
a
il
:
m
oh
_h
a
t84@y
ah
oo.com
1.
INTROD
U
CTION
Pr
ese
ntly
,
m
ultim
edia
network
a
nd
com
m
un
ic
at
ion
is
s
ha
rp
ly
pro
gressi
ng.
Di
gital
com
m
un
ic
at
ion
has
bec
om
e
so
sign
ific
ant
f
or
m
any
or
ga
nizat
ion
s.
It
s
upports
eve
n
daily
li
fe
act
ivities
and
c
omm
un
ic
at
ion
s
[1,
2].
F
or
i
ns
ta
nce
,
we
t
ran
s
fer
m
edical
i
m
ages,
dig
i
ta
l
au
dio
s,
m
us
ic
,
vid
eo
s,
dig
it
al
book
s,
et
c.
This
tra
ns
m
is
sion
occ
urs
on
public
networks
s
uc
h
a
s
the
i
nter
net
w
hich
m
eans,
a
nyone
ca
n
easi
ly
acce
ss
them
.
As
a
res
ult,
tho
s
e
file
s
m
ig
ht
be
sim
ply
forg
e
d
or
m
anipu
la
te
d.
To
m
ini
m
iz
e
this
threat
,
inf
or
m
at
i
on
secur
it
y
has
com
e
to
protect
data
against
so
m
e
issues,
s
uch
a
s
piracy,
data
f
orger
y,
et
c
[
3,
4].
I
n
this
fiel
d,
data
hid
i
ng
act
s
as
a
rece
nt
a
nd
well
-
kn
own
re
search
to
pic
w
hich
is
sp
e
ci
fied
int
o
t
he
f
ol
lowing
top
ic
s
su
c
h
a
s
ste
ganogra
phy
that
is
con
cer
n
in
g
m
uch
to
protect
the
com
m
un
ic
at
ion
s
pri
vacy
[5
,
6].
Stegan
ogra
phy
aim
s
to
hid
e
a
secret
m
ess
age
us
in
g
a
ha
rm
le
ss
cov
er,
so
no
ot
her
par
ty
can
dete
ct
this
secur
e
d
m
essa
ge.
I
n
the
past
,
data
hid
in
g
s
ta
rted
in
co
nventional
te
ch
ni
qu
e
s
su
c
h
as
inv
isi
ble
ink
s
,
sp
rea
d
-
sp
ect
r
um
co
m
m
un
ic
at
ion
s,
a
nd
co
ver
t
c
ha
nnel
s
[
3,
7].
Cu
rr
e
ntly
,
this
ca
n
be
car
ried
ou
t
by
hid
i
ng
the
secret
inside
a
nothe
r
cov
e
r
file
[
8].
This
m
edium
m
igh
t
be
an
i
m
age,
a
ud
i
o,
vi
deo
,
or
te
xt
[
9
-
13]
.
Waterm
ark
i
ng
[14
-
16]
con
ce
al
s
secret
inform
ati
on
su
c
h
as
sign
at
ures
within
any
m
edia
without
hav
i
ng
a
pe
rc
eptual
distor
ti
on
[
17, 18].
Audio
data
hi
di
ng
sc
hem
es
are
cl
assifi
ed
int
o
tw
o
ty
pes:
lo
ssless
an
d
loss
y.
The
lossless
al
gorithm
consi
ders
retri
evin
g
both
th
e
secret
m
essage
an
d
the
ori
gin
al
co
ve
r
file
.
Loss
y
m
et
hod
cares
only
on
recoveri
ng
t
he
secret
inf
orm
at
ion
[19].
Ther
e
a
re
m
a
ny
pa
ram
et
ers
wh
ic
h
are
use
d
to
e
valua
te
the
perform
ance
of
au
dio
ste
ganogr
ph
y
sc
hem
e
s.
N
ot
al
l
of
th
e
m
sh
ould
be
m
et
,
since
each
im
ple
m
entati
on
m
ay
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.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
147
–
154
148
hav
e
diff
e
re
nt
pur
po
se
s.
W
e
can
su
m
m
ariz
e
so
m
e
sign
ific
ant
on
es
in
F
ig
ure
1
,
w
hich
can
be
descr
i
bed
a
s
fo
ll
ows
[
11, 20]
:
a)
Ca
pacit
y
[21]
: t
his r
e
pr
ese
nts
the am
ou
nt
of the sec
ret m
ess
age th
at
can
b
e
carr
ie
d by the
cov
e
r fil
e.
b)
Im
per
cepti
bili
ty
(steg
o
file
qual
it
y):
it
con
s
iders
t
he
qu
al
it
y
of
the
ste
go
aud
i
o
file
after
bein
g
m
erg
e
d
with the
secret
m
essage.
c)
Secrecy
(Tem
per
co
nfr
on
ta
ti
o
n
):
this
m
eans
that
the
co
ncea
le
d
m
essage
is
secur
e
d
a
nd
it
is
diff
ic
ult
f
or
un
a
utho
rized
pa
rtie
s
to
acce
s
s
it
.
As
an
ex
a
m
ple,
this
ca
n
be
a
secret
key
that
is
s
ha
red
bet
ween
th
e
sen
der
a
nd the
receiver
.
d)
Com
pu
ta
ti
on
al
com
plexity
:
it
con
ce
rn
s
th
e
processi
ng
tim
e
ta
ken
by
e
m
bed
di
ng
an
d
ext
racti
on
op
e
rati
ons,
w
hich
s
houl
d
not
be
lo
ng.
Especial
ly
,
this
app
li
es
to
const
raine
d
de
vices,
s
uch
as
sm
artph
on
e
.
e)
Robus
tne
ss:
this
enab
le
s
us
f
or
rec
overin
g
the
secret
data
from
the
ste
go
aud
io
file
eve
n
if
it
is
being
at
ta
cked
.
Figure
1
.
Stega
nograp
hy m
eas
ur
em
ent p
ar
am
et
ers
In
a
udio
data
hid
in
g
syst
em
s
,
it
is
essenti
al
ly
req
uire
d
to
hav
e
high
payl
oad
w
hile
m
ain
ta
inin
g
th
e
qu
al
it
y,
a
nd
at
the
sam
e
tim
e
t
he
e
xistence
of
the
m
essage
s
hould
not
be
e
asi
ly
detect
ed.
Th
us
,
m
any
re
searc
h
pap
e
rs
a
re
sti
ll
lookin
g
f
orwa
rd
t
o
accom
plish
these
iss
ues
.
This
is
because
HAS
(
H
uma
n
A
udit
or
y
S
yst
e
m
)
is
an
obsta
cl
e
wh
ic
h
sta
nds
against
data
hid
in
g
al
gorit
hm
s
since
it
has
a
wide
dy
na
m
ic
and
dif
fere
ntial
range
[22]
.
Audio
data
hi
ding
te
ch
nique
s
us
in
g
Least
Sign
i
ficant
Bi
t
(LSB)
s
ubsti
tuti
on
te
c
hn
i
ques
[5]
ha
ve
m
uch
of
sim
pl
ic
it
y
a
nd
popu
la
rity
.
They
ca
n
ha
ve
a
great
pe
rfor
m
ance
i
n
te
rm
s
of
the
e
m
bed
di
ng
ca
pacit
y
and
com
pu
ta
ti
on
al
c
om
plexity
.
On
the
ot
her
ha
nd,
t
heir
sec
ur
it
y
le
ve
l
against
hac
ker
s
is
lo
w
a
nd
no
t
com
par
able.
T
he
sec
ret
m
essage
ca
n
be
det
ect
ed
a
nd
eve
n
obta
ine
d
easi
l
y
.
So
m
e
resear
cher
s
ha
ve
so
l
ved
this
pro
blem
by
com
bin
ing
LSB
ste
gano
gr
a
ph
y
with
a
crypto
gr
a
phic
schem
e
in
order
to
i
ncr
ease
t
he
se
cur
it
y
le
vel.
H
ow
e
ve
r,
this
is
no
t
a
sat
isfact
or
y
so
luti
on
du
e
t
o
the
wasted
com
pu
ta
ti
on
al
com
plexity
a
nd
l
os
t
band
width st
or
a
ge.
In
this
pap
e
r,
we
pro
pose
a
new
sc
hem
e
t
hat
ex
plo
re
s
di
gital
op
erat
or
s
(e.g.
X
OR
&
NO
T
gates
)
m
erg
ed
by
an
LSB
substi
tuti
on
te
c
hn
i
qu
e
.
This
resea
rch
i
s
app
li
ed
i
n
the
tem
po
ral
dom
ai
n,
the
au
di
o
cov
e
r
is
extracte
d
int
o
sam
ples
and
are
conver
te
d
into
bin
a
ry.
I
n
this
researc
h,
eac
h
sam
ple
is
rep
rese
nted
by
16
bits.
In
the
em
bed
di
ng
process,
the
po
sit
ions
of
t
wo
ra
ndom
bit
s
are
sel
ect
ed
within
the
inte
rv
al
(
5,
16)
as
a
key.
Accor
ding
to
that
posit
ion
,
t
wo
bits
from
e
ach
sam
ple
are
sel
ect
ed
to
be
us
e
d
as
a
g
uid
e
that
helps
rec
overi
ng
the
secret
m
es
sage.
T
he
em
b
edd
i
ng
is
pe
rfor
m
ed
by
app
l
yi
ng
X
OR
gate
on
the
secret
bits
and
the
sel
ect
ed
sam
ple’s
bits.
The
f
our
X
OR
ou
t
pu
ts
a
re
c
om
par
ed
with
t
heir
c
orrespo
ndin
g
sam
ple’s
le
ast
sign
ific
an
t
bits.
If
they
are
not
m
at
ched
then
the
LSBs
are
change
d
by
us
i
ng
a
n
inv
e
rter
(NOT)
lo
gical
op
e
rato
r.
Ot
he
rw
ise
,
there is
no c
ha
ng
e
. No
rm
al
l
y, this
helps
i
ncrea
sing t
he dif
ficult
y of detect
ing t
he se
cret m
essage.
The
re
st
of
t
hi
s
pa
per
is
str
uc
ture
d
as
the
f
ollow
i
ng.
I
n
S
ect
ion
2,
we
de
scribe
s
om
e
e
xisti
ng
a
udio
ste
gano
gr
a
ph
y
schem
es,
and
determ
ine
their
streng
t
hs
an
d
weaknesse
s.
T
he
detai
l
of
the
pr
op
os
e
d
m
e
t
hod
is
pro
vid
e
d
in
Se
ct
ion
3.
Sect
io
n 4 prese
nts t
he
experim
ental
res
ults. T
he
c
oncl
us
i
on
s
are
draw
n
i
n
Sect
io
n 5.
2.
RE
LAT
ED
RE
SEARCH
This
sect
ion
pro
vid
es
t
he
ba
sic
theor
y
ab
out
so
m
e
aud
io
data
hid
in
g
m
et
ho
dolo
gies
as
fo
ll
ows.
Audio
data
hi
ding
resea
rch
has
bee
n
popula
r
due
to
it
s
stren
gth
,
for
exam
ple
it
s
si
m
pl
ic
it
y.
Gen
erall
y,
aud
i
o
am
pli
tud
e
values
a
re
represe
nted
in
rea
l
nu
m
ber
s,
so
durin
g
the
e
m
bed
ding
th
e
secret
bits
do
es
not
cause
m
uch
noise
[23]
.
It
is
norm
al
that
e
ach
schem
e
has
it
s
ow
n
pro
s
and
c
ons.
A
ll
are
m
easur
ed
an
d
evaluate
d usin
g
the
af
or
em
en
ti
on
ed
d
a
ta
h
i
di
ng
c
rite
rias.
Bi
nn
y
et
al
.
[24]
has
pr
opos
e
d
a
m
et
ho
d
tha
t
us
es
LSB
of
each
sam
ple
to
hid
e
the
sec
ret
inform
at
ion
wh
il
e
m
ai
ntain
ing
the
ste
go
file
i
m
per
cepti
bili
ty
.
Nev
erthel
ess,
they
fail
to
get
the
secrec
y
and
the
r
obust
n
es
s
R
o
b
u
s
t
n
e
s
s
Q
u
a
l
i
t
y
S
e
c
u
r
i
t
y
C
a
p
a
c
i
t
y
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
Develo
ping
audio
data
hid
i
ng sch
e
me usin
g rand
om s
a
m
pl
e b
it
s wi
th
…
(
Mo
hamm
e
d H
atem Al
i Al
-
H
ooti
)
149
and
c
ould
not
fu
ll
y
rec
ov
e
r
the
ori
gin
al
a
ud
i
o
file
.
In
t
he
ne
xt
re
sear
ch,
Al
-
Hooti
et
al
.
[20]
des
ign
a
n
al
gorithm
bas
ed
on
m
od
ulus
f
un
ct
io
n
a
nd
locat
ion
m
aps.
T
his
af
fects
the
qu
al
it
y
and
the
rev
e
rsi
bili
ty
,
wh
ic
h
a
re succ
essfu
ll
y
ob
ta
in
ed.
I
n
c
on
tra
st,
this
work is f
r
agile
to
sta
nd a
gainst
hac
ker
s
’
d
et
ect
io
n.
Chowdh
ur
y
et
al
.
[
25]
ha
s
w
orke
d
on
a
m
et
ho
d
base
d
on
LSB
an
d
it
is
c
om
bin
ed
wit
h
a
crypto
gr
a
phic
schem
e
called
patte
rn
m
at
chi
ng
al
go
rithm
.
This
is
done
in
or
de
r
to
enc
rypt
the
secret
te
xt
so
it
s
secrecy
is
gu
a
ran
te
e
d.
By
usi
ng
a
sim
i
la
r
idea,
i
n
[26]
th
e
auth
ors
pro
pose
a
n
al
gorithm
that
i
m
ple
m
ents
LSB
-
based
dat
a
hid
i
ng
in
a
n
andr
oid
e
nviro
nm
ent.
This
sc
hem
e
sta
rts
encr
ypti
ng
the
se
cret
m
essage,
a
nd
put
it
in
MP3/WAV
au
dio
.
Me
a
nwhile
,
the
c
over
au
dio
is
divi
ded
int
o
re
gions.
Af
te
r
that,
the
secret
m
es
sage
is
e
m
bed
de
d by
m
od
ify
ing
thes
e re
gions. H
owever, the
em
bed
di
ng cons
um
e
s m
uch
tim
e to pro
ces
s.
Au
t
hors
of
[
21
]
has
c
ombine
d
Ge
ner
al
iz
ed
Diff
e
re
nc
e
Exp
a
ns
io
n
and
Re
duce
d
Dif
fer
e
nce
Ex
pan
si
on
to
pro
du
ce
a
ne
w
schem
e
that
i
s
able
to
hid
e
m
or
e
payl
oad
and
m
ai
ntain
t
he
qual
it
y.
Howev
e
r,
this
work
do
e
s
no
t
pay
at
te
ntion
t
o
the
secrecy
.
Thi
s
m
et
ho
d
ha
s
increase
d
th
e
capaci
ty
,
bu
t
their
i
m
per
cepti
bili
ty
m
ay
n
ot b
e
hi
gh
e
noug
h.
The
m
et
ho
d
de
scribe
d
in
[
27]
has
propose
d
a
rev
ers
i
ble
ste
gano
gr
a
phic
schem
e
us
ing
pro
ba
bili
sti
c
DNA
-
X
OR
secret
sh
ari
ng
-
ba
sed
col
or
im
ag
e.
Si
m
il
ar
to
ot
her
res
earc
h,
e
ach
D
NA
-
X
O
R
truth
ta
ble
va
lue
is
evaluate
d
by
usi
ng
PS
NR.
T
he
input
that
ha
s
the
hig
he
st
PSN
R
val
ue
is
con
sid
ere
d
for
the
secret
sha
rin
g.
T
hese
i
nputs
a
re
hi
dd
e
n
i
n
th
e
co
ver
im
age
wh
ic
h
are
us
e
d
as
th
e
key.
This
w
ork
di
vid
es
the
secret
sh
are
s
into th
ree
gro
ups: R
, G,
B wh
ere the
secret
m
essage is em
bedde
d.
As
it
is
m
entio
ne
d,
we
f
oc
us
on
the
te
m
po
ral
dom
ai
n
and
L
SB
au
dio
data
hid
in
g
re
pr
ese
nts
hi
gh
per
ce
ntage
fro
m
the
cur
re
ntly
published
m
et
hods
in
this
do
m
ai
n.
Mor
eov
e
r,
L
SB
-
ba
sed
m
et
ho
ds
s
i
m
ply
increase
or
dec
rease
one
valu
e
fr
om
the
LSB
value
of
eac
h
sam
pl
e
[28]
.
Howe
ver,
thes
e
schem
es
are
no
t
abl
e
to
f
ulfill
the
ot
her
ste
ga
nograph
ic
re
qu
i
re
m
ents
su
c
h
as
secrecy
a
nd
r
obus
t
ness.
Als
o,
t
he
sec
ret
m
essage
m
igh
t be l
os
t e
asi
ly
if th
e steg
o fil
e b
ei
ng att
acked o
r
c
om
pr
esse
d
[
23]
.
3.
RES
EA
R
CH MENTH
OD
This
sect
io
n
e
xp
li
ci
tl
y
descr
i
bes
th
e
pr
opose
d
al
go
rithm
of
this
resea
rch.
Com
m
on
ly
,
LSB
-
base
d
e
m
bed
di
ng
m
e
thods
ha
ve
bee
n
pro
po
s
ed
in
m
any
researc
h
pa
pers
[24]
.
H
ow
e
ve
r,
t
he
re
searche
rs
do
not
pay
m
uch
at
te
ntion
to
t
he
sec
recy (Tem
per
co
nfr
on
ta
ti
on)
feat
ure. A
s a
re
su
lt
, th
os
e
m
et
ho
ds
are
le
ss
us
e
fu
l
in
the
current
pract
ic
al
data
hid
ing
app
li
cat
ion
s
.
This
is
because
of
the
secr
et
data
extracti
on
sim
pl
icity
an
d
detect
abili
ty
[24]
.
Currentl
y,
secur
e
d
LSB
s
chem
es
are
only
the
on
es
w
hi
ch
are
com
bin
ed
with
one
of
the
encr
y
ption
w
orks
[
25]
.
This
c
os
ts
m
uch
in
the
com
pu
ta
ti
onal
com
plexity
.
In
acc
orda
nce
,
we
m
ai
nly
fo
cus
on
increasin
g
t
he se
cur
it
y pa
ram
et
er.
Norm
al
l
y,
com
bin
ing
data
hid
in
g
wit
h
cr
ypto
syst
e
m
[26]
can
inc
rea
se
the
secur
it
y
le
vel
bu
t
it
requires
m
uch
com
pu
ta
ti
on
al
com
plexity
.
Acco
r
ding
t
o
thi
s
co
nd
it
io
n,
w
e
desig
n
a
n
L
SB
-
bas
ed
a
udio
data
hid
in
g
m
et
ho
d
by
co
ns
i
de
rin
g
the
payl
oad,
qual
it
y,
secrecy
,
an
d
com
pu
ta
ti
on
al
com
plexity
aspect
s.
This
w
ork
doe
s
not
pay
at
te
nt
ion
to
the
rob
ust
ness.
Dif
fer
e
nt
from
the
pr
e
vious
re
searc
h
[22,
24]
,
we
propose
a
new
te
c
hn
i
que
that
us
e
s
tw
o
lo
gical
op
e
ra
tors
(
X
OR
an
d
NO
T
)
ba
sed
on
LSB
s
ub
sti
tu
ti
on
.
T
his
al
go
rith
m
m
ai
nly
hid
es
f
our
bits
in
eac
h
sam
ple
w
hich
is
pr
e
viously
co
nv
e
rted
into
16
bits.
T
hese
bits
are
di
vid
e
d
in
t
o
two
par
ts
as
i
ndic
at
ed
in
Fig
ure
2.
The
first
par
t
is
use
d
t
o
carry
the
ou
t
pu
ts
of
eac
h
X
O
R
op
e
rati
on,
a
nd
t
he
rand
om
two
se
le
ct
ed
bits a
re
chosen
in
t
he
s
econd
pa
rt that
contai
ns
12
bits.
Figure
2
.
T
he
us
e
of au
dio sa
m
ple r
epr
ese
nted
in
b
it
s
3.1.
The
Proc
edure
of Em
beddin
g
The
em
bed
ding
operati
on
is
il
lustrate
d
as
sh
ow
n
in
Fig
ure
3
and
as
sum
m
arized
in
t
he
ne
xt
fiv
e
ste
ps
as
foll
ows:
Step
1:
read
the
sec
ret
m
essage
a
nd
c
onsider
each
f
our
bits
(
,
1
,
,
2
,
,
3
,
,
4
)
to
be
e
m
bed
de
d
within
eac
h sa
m
ple.
Step
2:
rea
d
t
he
sam
ples
of
a
ud
i
o
file
(excl
ud
i
ng
the
first
40
sam
ples)
w
her
e
each
sam
ple
is
in
th
e
interval
[
0,
65535],
an
d
c
onve
rt
eac
h
sam
ple
into
16
bits.
Ever
y
sam
ple
is
div
ide
d
int
o
t
wo
par
ts
.
The
f
irst
1
1
1
1
1
1
1
0
1
1
0
1
0
0
0
1
L
S
B
M
S
B
s
e
l
e
c
t
i
o
n
p
a
r
t
e
m
b
e
d
d
i
n
g
p
a
r
t
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.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
147
–
154
150
on
e
c
onta
ins
4
bits
w
hich
a
r
e
us
e
d
to
be
c
ha
ng
e
d.
The
seco
nd
pa
rt
co
ntain
s
12
bits
that
t
wo
of
them
,
a
nd
,
are
sel
ect
ed
ra
ndom
ly
to
be
proces
sed
with
t
he
first
a
nd
t
he
thir
d
sec
ret
bits
us
i
ng
XO
R
ga
te
(d
e
note
d
in
this
pa
pe
r
by
).
It
is
worth
to
note
that
t
he
po
sit
ion
of
these
two
c
hose
n
bits
are
sent
to
t
he
destinat
io
n
a
s
a
key.
Step
3:
co
ns
i
de
r
eac
h
f
our
s
ecret
bits
a
nd
app
ly
the
X
O
R
lo
gical
op
e
r
at
or
as
s
how
n
in
(1).
Her
e
,
,
1
is
the
outp
ut
of
the
fi
rst
X
OR
-
in
g
that
is
ap
pl
ie
d
betwee
n
t
he
fir
st
secret
bit
and
t
he
fi
rs
t
sel
ect
ed
sam
pl
e
bin
a
ry
dig
it
w
her
e
is
an
in
de
x
that
re
pr
e
se
nts
the
num
ber
of
the
us
e
d
sa
m
ples.
Both
the
two
values
and
are
in
the r
an
ge
of
5
an
d
16
w
hic
h
de
note
s
the
posit
ion
o
f
the
ra
ndom
selected
bits
,
an
d
,
,
resp
ect
ively
.
It
is
com
pu
lso
rily
that
bo
th
r
andom
po
sit
io
n
val
ues
sho
uld
not
be
the
s
a
m
e.
The
pa
ra
m
et
er
,
2
carries
the
ou
t
pu
t
val
ue
of
XO
R
-
in
g
the
seco
nd
sec
ret
bit
,
2
an
d
t
he
f
irst
outp
ut
,
1
.
To
increa
se
the
c
om
plexity
of
com
pr
om
isi
ng
the
m
et
ho
d,
t
he
sec
ond
rando
m
sa
m
ple
bit
,
is
X
OR
-
e
d
with
the
t
hir
d
secret
bit
,
3
.
Af
te
r
that,
we
i
m
ple
m
ent X
O
R t
o
the
outp
ut
,
3
with
fo
ur
th
se
cret bit
,
4
.
,
1
=
,
1
,
,
2
=
,
2
,
1
,
3
=
,
3
,
,
4
=
,
4
,
3
(1)
Step
4:
em
bed
al
l
the
ou
tp
ut
s
,
1
,
,
2
,
,
3
,
and
,
4
into
e
ver
y
sam
ple
that
is
locat
ed
in
the
first
par
t.
Ba
sed
on
(
2)
,
the
outp
ut
s
are
c
om
par
e
d
with
eac
h
fir
st
par
t
of
sam
ple
bit
,
1
,
,
2
,
,
3
,
,
4
.
If
the
y
are
no
t e
qual
, t
hen each
sam
ple b
it
is flipp
e
d usi
ng NOT l
ogic
al
o
pe
rato
r.
Othe
rw
ise
,
t
he
re is
no c
hange.
,
1
=
{
,
1
=
(
,
1
)
,
,
1
=
,
1
,
1
=
,
1
,
ℎ
,
2
=
{
,
2
=
(
,
2
)
,
,
2
=
,
2
,
2
=
,
2
,
ℎ
,
3
=
{
,
3
=
(
,
3
)
,
,
3
=
,
3
,
3
=
,
3
,
ℎ
,
4
=
{
,
4
=
(
,
4
)
,
,
4
=
,
4
,
4
=
,
4
,
ℎ
}
(2
)
Step
5:
buil
d
the
ste
go
aud
i
o
file
based
on
the
new
m
od
if
ie
d
sam
ple
val
ues.
T
hen
it
is
sent
to
the
receiver
v
ia
any
p
ubli
c n
et
work.
As
a
n
exam
ple
for
this
em
beddin
g
pr
ocedur
e
,
assum
e
we
ha
ve
6
a
udio
sam
ples
w
hich
a
re
conve
rted
into
bin
a
ry
as
pr
ese
nted
in
Fi
gure
4.
Th
ese
t
wo
sam
ples
ar
e
sub
div
ide
d
i
nt
o
tw
o
par
ts
,
nam
el
y
the
e
m
bed
di
ng
par
t
wh
ic
h
is
r
epr
ese
nted
by
four
le
ast
si
gn
i
ficant
b
it
s,
a
nd
the
sec
ond
pa
rt
that
is
us
e
d
to
ge
ner
at
e
t
he
two
rand
om
l
y
cho
s
en
bits.
S
uppos
e
the
sen
der
c
hoose
s
the
bit
nu
m
ber
5
as
fir
s
t
po
sit
io
n
an
d
bit
nu
m
ber
10
as
the
seco
nd
posit
ion
.
T
he
n,
base
d
on
(1)
an
d
(2)
the
ob
ta
i
ne
d
res
ults
are
pr
ese
nted
i
n
the
Tab
le
1
w
her
ea
s
su
ccess
fu
ll
y
the
ste
go
f
our
LSB
bits
are
rep
la
ced
by
the
f
our
outp
ut
s
(
1
,
2
,
3
,
4
)
of
the
E
xclusi
ve
OR ope
rati
on.
Figure
3
.
Th
e
e
m
bed
din
g p
r
oc
ess
T
r
a
n
s
f
o
r
m
i
n
b
i
n
a
r
y
f
o
r
m
m
e
s
s
a
g
e
c
o
v
e
r
O
b
t
a
i
n
e
a
c
h
f
o
u
r
b
i
t
s
t
o
b
e
e
m
b
e
d
d
e
d
C
o
n
v
e
r
t
i
t
i
n
b
i
n
a
r
y
S
e
g
m
e
n
t
t
h
e
s
a
m
p
l
e
i
n
t
o
2
p
a
r
t
s
G
e
n
e
r
a
t
e
t
w
o
r
a
n
d
o
m
v
a
l
u
e
s
P
r
o
c
e
s
s
t
h
e
e
m
b
e
d
d
i
n
g
u
s
i
n
g
X
O
R
a
n
d
N
O
T
g
a
t
e
s
B
u
i
l
d
t
h
e
s
t
e
g
o
a
u
d
i
o
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
Develo
ping
audio
data
hid
i
ng sch
e
me usin
g rand
om s
a
m
pl
e b
it
s wi
th
…
(
Mo
hamm
e
d H
atem Al
i Al
-
H
ooti
)
151
Figure
4
.
Ex
a
m
ple o
f
the
em
beddin
g p
ro
ces
s
Table
1
.
T
he
E
m
bed
din
g E
xa
m
ple O
utp
uts
A
b
1
O
1
b
2
O
2
b
3
B
O
3
b
4
O
4
1
0
1
1
0
0
1
1
0
1
1
1
0
0
0
0
1
1
0
1
0
0
0
1
1
1
0
1
0
1
1
1
0
0
0
0
0
0
1
1
3.2.
The
Ex
tr
act
i
on
Pr
oced
ure
It
is
com
m
on
that
the
extrac
ti
on
pr
ocess
is
j
us
t
i
nv
e
rsing
the
em
bed
ding
ope
rati
on.
T
he
recei
ver
ta
kes
the
deliv
ered
ste
go
file
an
d
tw
o
c
hos
en
key
val
ues
that
in
dicat
es
the
posit
io
ns
of
t
he
tw
o
sel
ect
ed
sam
ple b
it
s.
Aft
er th
at
, t
his
procedu
re is
perf
or
m
ed
as s
how
n
in
Fig
ure
5
a
nd the
f
ollow
i
ng ste
ps
:
Step
1:
ext
ract
the
ste
go
au
di
o
file
into
s
am
ples
wh
ic
h
are
co
nv
e
rted
into
bin
a
ry.
C
orres
pond
to
th
e
e
m
bed
di
ng pr
oc
ess, the
f
ir
st 4
0
sam
ples are
i
gnored
.
Step
2:
inse
rt
the
tw
o
key
va
lues
,
into
the
extracti
on
syst
e
m
to
determ
i
ne
the
po
sit
io
ns
of
th
e
rand
om
l
y sel
ect
ed
sam
ple two bits
,
′
,
,
′
.
Step
3:
a
pp
ly
XO
R
oper
at
or
betwee
n
t
he
fir
st
sel
ect
ed
sam
ple
bit
,
′
a
nd
the
first
LSB
of
t
he
ste
go
sam
ple.
As
a
r
esults,
t
he
first
secret
bit
,
1
′
is
obta
ined
.
This
f
ir
st
secret
bit
is
al
so
XO
R
-
ed
with
the
se
c
on
d
LSB
of
t
he
ste
go
sam
ple
in
order
t
o
obta
in
the
seco
nd
se
cret
bit
,
2
′
.
A
fter
that,
the
thir
d
secret
bit
,
3
′
is
ob
ta
ine
d by X
OR
-
in
g
the
thi
r
d
LSB
of
t
he
s
a
m
e
sam
ple
,
3
′
a
nd
the
sec
ond sel
ect
ed
rand
om
b
inary d
igit
,
′
.
Finall
y,
getti
ng
the
f
ourth
sec
r
et
bin
ary
dig
it
,
4
′
XO
R
is
a
pp
li
e
d
to
th
e
obta
in
ed
thir
d
sec
ret
dig
it
,
3
′
an
d
th
e
sam
ple f
ourth LSB
,
4
′
as
in E
qu
at
ion
(3).
Figure
5
.
T
he
e
xtracti
on
op
e
ra
ti
on
,
1
′
=
,
1
′
,
′
,
2
′
=
,
1
′
,
2
′
,
3
′
=
,
3
′
,
′
,
4
′
=
,
3
′
,
4
′
(3)
H
i
1
1
1
1
1
1
1
0
1
1
0
1
1
1
0
1
1
1
1
1
1
1
1
1
0
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
0
0
0
0
0
0
0
0
0
1
0
0
1
1
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
1
1
1
0
0
0
0
0
0
0
1
0
1
1
0
1
0
0
0
L
S
B
M
S
B
S
e
c
r
e
t
m
e
s
s
a
g
e
0
1
1
0
1
0
0
1
6
5
2
3
5
6
5
4
0
1
4
1
5
7
2
6
3
3
6
0
A
u
d
i
o
s
a
m
p
l
e
s
A
u
d
i
o
s
a
m
p
l
e
s
i
n
b
i
n
a
r
y
0
1
0
0
1
0
0
0
O
b
t
a
i
n
t
h
e
s
e
c
r
e
t
m
e
s
s
a
g
e
a
n
d
t
r
a
n
s
f
o
r
m
i
t
i
n
t
o
i
t
s
o
r
i
g
i
n
a
l
f
o
r
m
m
e
s
s
a
g
e
S
t
e
g
o
C
o
n
v
e
r
t
i
t
i
n
b
i
n
a
r
y
S
e
g
m
e
n
t
t
h
e
s
a
m
p
l
e
i
n
t
o
2
p
a
r
t
s
P
r
o
c
e
s
s
t
h
e
e
x
t
r
a
c
t
i
o
n
o
p
e
r
a
t
i
o
n
I
n
s
e
r
t
t
h
e
t
w
o
r
a
n
d
o
m
p
o
s
i
t
i
o
n
s
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.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
147
–
154
152
Step
4:
finall
y,
com
bin
e
al
l
of
the
ob
ta
ine
d
s
ecret
bits
,
1
′
,
,
2
′
,
,
3
′
,
and
,
4
′
wh
ic
h
a
re
obta
ined
f
ro
m
eac
h
sam
ple in
a v
ect
or an
d
c
onve
rt that
vec
tor
i
nto
it
s
or
i
gi
nal form
that m
igh
t be te
xt, a
ud
i
o,
or im
age.
4.
RESU
LT
S
A
ND AN
ALYSIS
In
t
his
re
searc
h,
t
he
sec
ret
m
essage
is
ge
ner
at
e
d
us
in
g
Ra
nd
i
functi
on
in
(Mat
la
b
-
2017
a
).
For
t
he
aud
i
o
co
ver,
we
us
e
10
a
udio
*
.w
a
v
file
s.
Each
co
ve
r
file
has
it
s
own
du
rati
on
and
siz
e,
ho
wev
e
r,
al
l
are
sa
m
ple
d
in
16
bit
dept
h
[29
-
31]
.
It
i
s
intende
d
that
the
file
s
are
c
ho
s
en
in
dif
fe
r
ent
cases
e.g
.
s
peech,
m
us
ic
.
So
m
e
a
re
in
high
le
vel
and
ot
her
s
has
low
s
ound
le
ve
l.
The
nu
m
ber
of
sam
ples
(excl
ud
in
g
t
he
40
on
e
s
wh
ic
h
ar
e
ig
nored
since
t
hey
can
be
im
pacte
d)
is
va
ried.
T
his
sim
ula
ti
on
is
done
i
n
a
c
om
pu
te
r
la
ptop
Asus
that
has
co
re
i3
proces
sor,
6
GB
RAM,
an
d
500
GB
H
D
.
Four
sce
nar
i
os
are
c
onduct
ed
to
eval
uate
fou
r
featur
e
s as
f
ollow
s:
First, ca
pacit
y or
em
beddin
g
r
at
e that i
s r
ep
re
sented
a
s s
how
n
in
Table
2.
T
his m
et
ho
d
is a
ble to
hid
e
secret
bits
4
ti
m
es
the
num
ber
of
t
he
co
ve
r
sam
ples
wh
ic
h
m
eans
each
sa
m
ple
is
able
to
carry
4
bits
w
her
ei
n
the
existi
ng
w
orks
[20,
24]
e
ach
sam
ple
can
on
ly
car
ry
one
bit
.
H
ow
e
ve
r,
the
m
or
e
th
e
inserte
d
da
ta
causes
sli
gh
t i
m
pact o
n
the
quali
ty
o
f
the ste
go f
il
e a
s in
dicat
ed
in
both Ta
ble
3
a
nd Ta
ble
4.
Seco
nd,
im
per
cepti
bili
ty
wh
i
ch
is
m
easur
e
d
by
us
in
g
Si
gnal
to
N
oise
Ra
t
io
(SNR)
an
d
Peak
Sig
nal
to
No
ise
Ra
ti
o
(PSNR
)
as
in
E
quat
io
n
(4)
and
(
5),
res
pe
ct
ively
[8]
.
Th
e
va
riable
de
no
te
s
the
num
ber
of
sam
ples,
an
d
C
m
eans
the
c
over
a
udio
w
he
re
the
ste
go
file
is
pr
ese
nted
by
wh
il
e
µ
re
pr
ese
nt
s
the
m
axi
m
um
value
of
au
dio
sam
ple
.
The
hig
he
r
the
value
of
S
NR
and
P
SN
R,
the
great
er
the
si
m
il
arity
between
the
cov
e
r
and the
ste
go
fi
le
s.
=
∑
2
=
1
∑
(
−
)
2
=
1
(4)
=
µ
2
∑
(
−
)
2
=
1
(5)
Thir
d,
the
sec
ur
it
y
par
am
et
er
is
i
m
pr
ove
d
in
this
pa
per.
This
is
beca
use
in
m
os
t
exist
ing
m
od
ul
us
and
LSB
w
ork
s
su
c
h
as
[
20,
24
]
,
t
he
sec
ret
m
essage
rec
ov
ery
is
sim
ply
ob
ta
ined
by
ta
kin
g
the
LSB
of
each
ste
go
sam
ple.
I
n
co
ntrast,
t
his w
ork
hi
des
the
4
outp
uts
that are
obta
ine
d
f
r
om
XO
R
-
in
g
ga
te
s
in
each
sa
m
ple.
These
outp
uts
are
not
th
e
ex
a
ct
secret
bits.
Ther
e
f
or
e,
s
uppo
s
e
the
at
ta
c
ke
r
kn
ows
our
m
et
ho
dolo
gy,
he/she
needs
to
c
ondu
ct
16
t
rial
s
for
each
sam
ple
in
orde
r
to obtai
n
the
e
xact 4
s
ecret
bits.
In
t
hi
s
pap
e
r,
w
e d
e
sign
e
d
an
e
qu
at
io
n
t
ha
t
is
pr
ese
nted
in
(
6).
It
is
us
e
d
to
m
easur
e
the
am
ou
nt
of
tria
ls
that
are
re
qu
i
red
f
or
obta
inin
g
the
sec
ret
m
ess
age.
He
re,
is
t
he
nu
m
ber
of
t
i
m
es
need
e
d
t
o
il
le
gally
ob
ta
in
the
sec
ret
bit
s,
re
pr
ese
nts
t
he
four
sel
ect
ed
bi
ts,
is
the
pos
sibil
it
ie
s
of
ea
ch
sec
ret
bit
(
since
it
is
i
n
bi
nar
y
s
o
is
al
ways
2),
an
d
∁
is
sy
m
bo
li
zed as t
he
num
ber
o
f t
he
ste
go sam
ples that are em
bedde
d.
As
a r
esult, h
ac
kers
need
16 trial
s for
eac
h
sam
ple
in
or
de
r
to
obta
in
the
exact
f
our
secret
bits.
Th
is
is
m
ul
ti
plied
by
the
num
ber
of
the
em
bedd
e
d
sam
pl
es. Th
ere
fore,
the m
or
e
the am
ou
nt of
hidden sec
ret bi
ts t
he
higher
t
he
di
ff
ic
ulty
fo
r
hac
kers to
de
te
ct
it.
Ba
sed
on
E
qu
a
ti
on
(6),
le
t
the
em
bed
de
d
ste
go
sam
ples
am
ount
is
6615
00
an
d
it
is
know
n
in
eac
h
sam
ple
4
2
trai
ls
are
nee
de
d
in
orde
r
to
ob
ta
in
the
4
se
cret
bits,
s
o
t
he
possibil
it
y
of
il
le
gally
gaining
the
whole
s
ecret
m
essage
is
(
4
2
)
∗
661500
=
1
0
5
8
4
0
00
ti
m
es,
wh
e
reas
in
the
e
xisti
ng
wor
ks
[
20,
24]
al
low
a
nyone
t
o
directl
y extract
the sec
ret m
es
sage
by
ob
ta
ini
ng the
LSB
of
each sam
ple.
=
(
2
)
∗
(6)
Table
2
. T
he
E
m
bed
din
g R
at
e
base
d o
n
F
our
Scen
a
rios
No
File na
m
e
No
.
Sa
m
p
l
es
1
0
0
% p
ay
lo
ad
7
5
% p
ay
lo
ad
5
0
% p
ay
lo
ad
2
5
% p
ay
l
o
ad
1
0
0
1
_
0
1
7
5
2
0
6
8
3
0
0
8
2
7
2
2
2
5
6
2
0
4
1
5
0
4
1
3
6
7
5
2
0
6
8
2
0
0
1
_
0
2
8
1
5
5
2
0
3
2
6
2
0
8
0
2
4
4
6
5
6
0
1
6
3
1
0
4
0
8
1
5
5
2
0
3
AM2_
3
5
U_
5
5
7
8
3
3
7
7
2
3
1
3
3
5
0
8
1
7
3
5
0
1
3
1
1
1
5
6
6
7
5
4
5
7
8
3
3
7
7
4
Aco
m
ic
6
6
1
5
0
0
2
6
4
6
0
0
0
1
9
8
4
5
0
0
1
3
2
3
0
0
0
6
6
1
5
0
0
5
0
0
1
_
0
3
9
9
7
1
1
9
3
9
8
8
4
7
6
2
9
9
1
3
5
7
1
9
9
4
2
3
8
9
9
7
1
1
9
6
Blu
es
6
6
1
5
0
0
2
6
4
6
0
0
0
1
9
8
4
5
0
0
1
3
2
3
0
0
0
6
6
1
5
0
0
7
Clas
sical
6
6
1
5
0
0
2
6
4
6
0
0
0
1
9
8
4
5
0
0
1
3
2
3
0
0
0
6
6
1
5
0
0
8
0
2
.Scho
o
l Bo
y
-
9
1
7
6
3
9
9
9
7
0
5
5
9
9
6
5
2
9
1
9
9
7
3
5
2
7
9
9
8
1
7
6
3
9
9
9
9
Jazz
6
6
1
5
0
0
2
6
4
6
0
0
0
1
9
8
4
5
0
0
1
3
2
3
0
0
0
6
6
1
5
0
0
10
Ro
ck
6
6
1
5
0
0
2
6
4
6
0
0
0
1
9
8
4
5
0
0
1
3
2
3
0
0
0
6
6
1
5
0
0
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
Develo
ping
audio
data
hid
i
ng sch
e
me usin
g rand
om s
a
m
pl
e b
it
s wi
th
…
(
Mo
hamm
e
d H
atem Al
i Al
-
H
ooti
)
153
Table
3
. T
he
S
te
go
A
ud
i
o Quai
lt
y based on
Sign
al
t
o No
is
e Rat
io
No
File na
m
e
SNR v
alu
es in
dB
In1
0
0
%
In7
5
%
In5
0
%
In2
5
%
1
0
0
1
_
0
1
7
7
.
0
2
7
8
.27
8
0
.03
8
3
.02
2
0
0
1
_
0
2
7
6
.97
7
8
.25
8
0
.02
8
3
.03
3
AM2_
3
5
U_
5
7
6
.62
7
7
.86
7
9
.61
8
2
.58
4
Aco
m
ic
7
4
.46
7
5
.71
7
7
.47
8
0
.48
5
0
0
1
_
0
3
7
7
.02
7
8
.28
8
0
.05
8
3
.07
6
Blu
es
7
4
.05
7
5
.30
7
7
.06
8
0
.08
7
Clas
sical
7
4
.92
7
6
.17
7
7
.92
8
0
.94
8
0
2
.Scho
o
l Bo
y
-
9
7
6
.48
7
7
.72
7
9
.48
8
2
.49
9
Jazz
7
5
.94
7
7
.19
7
8
.96
8
1
.97
10
Ro
ck
7
6
.10
7
7
.35
7
9
.11
8
2
.12
Table
4
. T
he
S
te
go
A
ud
i
o Quai
lt
y based on
Peak Si
gnal
to Noise
Rat
io
No
File na
m
e
PSNR in
dB
In1
0
0
%
In7
5
%
In5
0
%
In2
5
%
1
0
0
1
_
0
1
8
0
.04
8
1
.29
8
3
.04
8
6
.04
2
0
0
1
_
0
2
7
9
.98
8
1
.26
8
3
.03
8
6
.04
3
AM2_
3
5
U_
5
8
0
.02
8
1
.27
8
3
.01
8
5
.99
4
Aco
m
ic
8
0
.05
8
1
.30
8
3
.06
8
6
.07
5
0
0
1
_
0
3
8
0
.02
8
1
.28
8
3
.05
8
6
.07
6
Blu
es
8
0
.04
8
1
.29
8
3
.04
8
6
.06
7
Clas
sical
8
0
.05
8
1
.30
8
3
.06
8
6
.08
8
0
2
.Scho
o
l Bo
y
-
9
8
0
.05
8
1
.29
8
3
.05
8
6
.06
9
Jazz
8
0
.0
5
8
1
.30
8
3
.06
8
6
.08
10
Ro
ck
8
0
.06
8
1
.31
8
3
.06
8
6
.07
Four
t
h,
c
om
pu
ta
ti
on
al
c
omplexit
y,
this
m
et
ho
d
c
os
ts
in
av
era
ge
near
ly
30.
20
seco
nd
s
to
be
i
m
ple
m
ented.
This
fast
com
puta
ti
on
al
proce
ss
al
lows
us
to
instal
l
the
pro
po
s
ed
syst
em
even
in
sm
artph
ones
.
B
ased
on
the
c
orrelat
ion
m
et
ho
d
[
20]
the
fu
ll
r
ec
ov
e
ry of t
he
secret m
essage is
100%
gua
ran
te
e
d
.
In
br
ie
f
,
this
work
has
s
uc
cessf
ully
ob
ta
ined
gr
eat
cap
aci
ty
and
m
ain
ta
ined
ste
go
fil
e
qu
al
it
y.
The
sec
recy
is
al
so
high
in
a
fast
com
pu
ta
ti
on
al
c
om
plexity
perform
ance.
In
dif
fere
nt
w
ords,
this
al
go
r
it
h
m
can
e
nh
a
nce
th
e
capaci
ty
,
sec
recy
an
d
c
om
plexity
wh
il
e
m
ai
ntainin
g
the
q
ualit
y.
O
n
t
he
oth
e
r
ha
nd,
thi
s
w
ork
is
no
t
rob
us
t
to
sta
nd
agai
ns
t
any
at
ta
cks
e.g
.
WAV
to
MP3
com
pr
ess
ion
.
T
heref
or
e
,
any
at
ta
ck
on
it
ca
n
cause l
os
in
g
th
e secret m
essage
wh
ic
h
ca
n b
e co
ns
tr
uctive i
n
s
om
e cases.
5.
CONCL
US
I
O
N
Audio
data
hidi
ng
has
it
s
secur
it
y
sign
ific
a
nc
e
in
m
any
field
s
su
c
h
as
bankin
g,
m
i
li
ta
ry,
and
ai
rlines
syst
e
m
s.
In
t
hi
s
pa
per,
we
present
a
s
chem
e
that
f
oc
us
es
o
n
im
pr
ov
in
g
the
ca
pacit
y,
com
plexity
,
an
d
the
secrecy
crit
eri
a
of
a
n
L
SB
su
bst
it
ution
al
gorithm
without
bei
ng
c
om
bin
ed
with
a
c
r
yptogra
ph
y
sc
hem
e
.
The
pro
ba
bili
t
y
of
detect
ion
is
m
ini
m
iz
e
d
by
us
ing
X
O
R
gates
wh
er
eas
the
com
pu
ta
ti
on
al
com
plexity
perform
ance
i
s
achieve
d.
I
n
the
w
or
st
case,
this
sch
e
m
e
con
su
m
es
around
30.
20
sec
onds
wh
e
rei
n
crypto
gr
a
phic
schem
es
can
consum
e
m
uc
h
of
c
om
pu
ta
ti
on
al
com
plexity
and
wast
e
m
uch
of
st
or
a
ge.
As
com
par
iso
n
bet
ween
our
m
et
ho
d
an
d
norm
al
LBS
wo
r
ks
we
a
re
a
ble
to
achie
ve
m
or
e
em
bed
di
ng
rate
wh
il
e
m
ai
ntaining
the
qual
it
y.
More
over
,
th
is
m
et
ho
d
ca
n
yi
el
d
a
gr
eat
secrecy
com
par
ed
t
o
the
e
xis
ti
ng
relat
ed
sc
hem
e
s.
I
n
the
f
uture
,
we
ar
e
plan
nin
g
to
inc
rease
the
em
bed
di
ng
rate
s
o
each
s
a
m
ple
can
car
r
y
one
byte
.
W
e
belie
ve
this
will
not
on
ly
increase
s
payl
oad
but
al
so
it
will
fu
r
ther
com
plica
t
e
the
secret
m
essage
detect
ion.
ACKN
OWLE
DGE
MENTS
The
a
uthors
w
ou
l
d
li
ke
to
e
xpress
t
heir
warm
est
gr
at
it
utes
to
Insti
tut
T
ek
no
l
og
i
Se
pulu
h
Nopem
ber
ITS, S
ur
a
baya,
Indo
nesia f
or the s
upport t
hat h
as
b
ee
n give
n
to
this
resear
ch.
REFERE
NCE
S
[1]
Bhui
y
an
SS
N,
et
al.
An
Im
prove
d
Im
age
Ste
ganogr
aph
y
Alg
orit
hm
base
d
on
PV
D.
Indone
sian
Journal
of
El
e
ct
ri
ca
l
Eng
in
ee
ring
and
Com
pute
r
Sc
ie
nc
e. 2
018;10:
569~77.
[2]
Din
R,
et
al.
A
Com
par
at
ive
Re
vie
w
on
Da
ta
Hi
ding
Schemes.
I
ndonesia
n
Journ
al
of
Elec
tr
ical
Engi
ne
eri
ng
an
d
Com
pute
r
Scie
n
ce
.
2018;11
:768
~74.
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.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
147
–
154
154
[3]
Al
-
Dm
our
H,
A
l
-
Ani
A.
A
steg
anogr
aph
y
embedding
m
et
hod
b
ase
d
on
edg
e
id
ent
ifica
ti
on
and
XO
R
codi
ng
.
Expe
rt
s
y
st
ems
with
Appli
ca
t
ion
s.
2016;46:
293
-
3
06.
[4]
Ahm
ad
T,
Fiqar
TP.
Enha
nc
in
g
the
Perform
anc
e
of
Audio
Data
Hiding
Method
b
y
Sm
oot
hing
Inte
rpol
at
e
d
Sam
ple
s.
Int
ern
a
ti
onal Journal of Innovati
v
e
Com
puti
ng,
In
form
ation
and
Control. 2018;14(3):757
-
79.
[5]
Begum
MB,
Venka
t
ara
m
ani
Y
.
LSB
base
d
audi
o
steg
anogr
aph
y
base
d
on
te
xt
compress
ion.
Proc
edi
a
Engi
ne
eri
ng.
20
12;30:
703
-
10.
[6]
Hus
sain
HS
,
et
al
.
The
Embeddin
g
Perform
anc
e
of
StegSVM
Mod
el
in
Im
age
Steg
anogr
aph
y
.
Indo
nesia
n
Journal
of
Elec
tr
ical Eng
ine
er
ing
and
Co
m
pute
r
Scie
n
ce. 2018;12:233~8.
[7]
Al
Huti
MH
A,
et
al.,
ed
it
ors.
In
c
rea
sing
th
e
ca
p
a
ci
t
y
of
th
e
sec
r
et
dat
a
using
DEpi
xel
s
bloc
ks
and
adj
usted
RDE
-
base
d
on
gra
y
sc
al
e
images
.
Inform
at
ion
&
Com
m
unic
at
ion
T
echnolog
y
and
S
ystems
(ICTS),
2
015
Inte
rna
t
ional
Confer
ence
on
; 2015:
IE
EE.
[8]
Mansor
FZ,
et
al
.
An
Anton
y
m
Subs
ti
tut
ion
-
base
d
Model
on
Li
nguisti
c
Ste
ganogr
aph
y
Me
t
hod.
Indone
sian
Journal
of El
ec
tr
ic
a
l
Eng
ineeri
ng
and
Com
puter S
ci
en
ce.
2018;12
:
225~32.
[9]
Devi
RR,
Pugaz
hent
hi
D.
Ide
al
Sam
pli
ng
Rat
e
t
o
Reduc
e
Distor
ti
on
in
Audio
Stega
nogra
ph
y
.
Proce
di
a
Com
pute
r
Scie
nc
e. 2016;
8
5:418
-
24.
[10]
Rahmani
P,
Dastgha
ib
y
far
d
G.
An
eff
ic
i
ent
histog
ram
-
base
d
inde
x
m
appi
ng
m
ec
hani
sm
for
rev
ersibl
e
data
hidi
ng
in
VQ
-
com
pre
ss
ed
images
.
Infor
m
at
ion
Sci
ences.
2017.
[11]
Shahadi
HI,
e
t
a
l.
Concurr
ent
ha
rdware
ar
chi
t
ec
t
ure
for
dual
-
m
o
de
audi
o
st
ega
n
ogra
ph
y
p
roc
ess
or
-
base
d
FP
GA
.
Com
pute
rs &
E
l
ec
tr
ic
a
l Engi
ne
er
ing.
2016
;49:
95
-
116.
[12]
Sloan
T,
Hern
ande
z
-
C
astro
J.
Forensic
an
al
y
sis
of
v
ide
o
stega
nogra
ph
y
t
ools.
Peer
J
Co
m
pute
r
Scie
n
ce.
2015;1:
e7
.
[13]
Yan
F,
et
a
l.
F
le
xible
r
epr
ese
n
ta
ti
on
and
m
anipulat
ion
of
aud
i
o
signal
s
on
qu
ant
um
computer
s.
The
or
et
i
cal
Com
pute
r
Scie
n
ce
.
2017.
[14]
Hu
H
-
T,
et
al.
S
y
nchr
onous
b
li
nd
a
udio
wa
term
ark
ing
via
shape
conf
igur
at
i
on
of
sorted
L
W
T
coe
ff
ic
i
ent
m
agni
tude
s.
Sig
nal
Proc
essing.
2
018;147:
190
-
20
2.
[15]
Hua
G,
et
al.
Twe
nt
y
y
ears
of
digi
tal
audi
o
wate
rm
ark
ing
—
a
comprehe
n
sive
rev
ie
w
.
Signal
Proce
ss
ing
.
2016;128:
222
-
4
2.
[16]
Le
i
BY,
et
al
.
Bl
ind
and
ro
bust
audi
o
wat
ermarki
ng
sche
m
e
base
d
on
SV
D
–
DCT.
Sig
nal
Proce
ss
in
g
.
2011;91(8):
1973
-
84.
[17]
Kar
DC,
Mulkey
CJ.
A
m
ult
i
-
th
reshold
base
d
a
udio
stega
nogra
ph
y
sche
m
e.
Jou
rna
l
of
information
sec
urity
and
appl
i
ca
t
ions.
20
15;23:
54
-
67.
[18]
Renz
a
D
,
L
em
us
C.
Authen
ti
c
ity
ver
ifica
ti
on
of
a
udio
signal
s
bas
ed
on
fra
g
il
e
w
a
te
rm
ark
ing
for
a
udio
fore
nsics
.
Expe
rt
S
y
st
ems
with
Appli
ca
t
ion
s.
2018;91:
211
-
2
2.
[19]
Chang
C
-
C.
A
r
eve
rsibl
e
data
h
idi
ng
sche
m
e
us
ing
complement
ar
y
embedding
strat
eg
y
.
Inform
at
ion
Sci
enc
es.
2010;180(16):
30
45
-
58.
[20]
Al
-
Hooti
A,
et
a
l.
Audio
Dat
a
Hi
ding
Based
on
Sam
ple
Value
Mo
difi
c
at
ion
Us
ing
Modulus
Functi
on.
Journal
of
Inform
at
ion
Pro
ce
ss
ing
S
y
st
ems
.
2016;12(3).
[21]
Andra
MB,
et
al
.
Med
ical
Re
cor
d
Protection
with
Im
prove
d
GRD
E
Data
Hi
ding
Method
on
Audio
Files
.
Engi
ne
eri
ng
Lett
ers.
2017;25
(2).
[22]
Li
X,
Yu
HH
,
edi
tors.
Tra
nsp
ar
ent
and
robust
a
udio
dat
a
hid
ing
in
subband
dom
ai
n.
Inform
at
i
on
Te
chnol
og
y
:
Coding
and
Co
m
puti
ng,
2000
Proce
ed
ings I
nt
er
nat
ion
al
Conf
erence
on
;
2000
:
IE
EE
.
[23]
Datt
a
B,
et
al.,
edi
to
rs.
Mult
i
-
bit
Dat
a
Hid
ing
in
Randoml
y
Chosen
LSB
L
a
y
ers
of
an
Au
dio.
Inform
ation
Te
chno
log
y
(ICI
T),
2016
Int
ern
a
t
iona
l
Confer
e
nc
e
on;
2016:
IEEE.
[24]
Binn
y
A,
Koilakuntla
M,
edi
to
rs.
Hiding
sec
ret
informati
on
using
LSB
bas
ed
audi
o
s
te
ga
nogra
ph
y
.
Soft
Com
puti
ng
and M
ac
hine Int
el
l
ig
enc
e
(ISCM
I),
2
014
Inte
rn
at
ion
a
l
Confer
ence
on;
2014:
I
EEE.
[25]
Chowdhur
y
R
,
et
al
.
A
vi
ew
o
n
LSB
base
d
a
udio
steg
anogr
a
ph
y
.
Int
ern
ation
al
Journa
l
of
S
ec
uri
t
y
and
Its
Applic
a
ti
ons.
20
16;10(2):
51
-
62.
[26]
Si
buria
n
R,
edit
or
Stega
nogra
ph
y
implement
at
io
n
on
andr
oid
smart
phone
using
t
he
LSB
(le
ast
si
gnifi
c
ant
bit
)
to
MP
3
and
W
AV
audi
o.
W
ire
l
ess
and
T
el
ema
ti
cs
(
ICW
T),
2017
3r
d
Inte
rn
at
ion
al C
onfe
ren
c
e
on
;
20
17:
IE
EE.
[27]
Tunc
er
T
,
Avci
E.
A
r
eve
rsib
le
d
at
a
hidi
n
g
al
gor
i
thm
base
d
on
pr
obabi
li
st
ic
DN
A
-
XO
R
sec
ret
sha
ring
sche
m
e
fo
r
col
or im
age
s.
Di
spla
y
s
.
2016;41
:
1
-
8.
[28]
Basu
PN
,
Bho
wm
ik
T,
edi
tors
.
On
embeddin
g
of
t
ext
in
au
dio
a
c
ase
of
stega
nogr
aph
y
.
Rec
en
t
Tr
ends
i
n
Inform
at
ion, T
e
l
ec
om
m
unic
at
ion
and
Com
puti
ng
(IT
C),
2010
Int
e
rna
ti
on
al
Conf
er
enc
e
on;
2010: I
EE
E
.
[29]
Tz
an
et
ak
is G. Dat
ase
ts
-
m
ars
y
as
[onli
n
e]
.
2002.
[30]
Urbana
-
Champa
ign
SS
TGUoIa.
Ill
ino
is
Speec
h
and
L
angua
g
e
Engi
ne
eri
ng.
Univer
sit
y
of
Illi
nois
at
Urban
a
-
Champaign;
200
3.
[31]
Bosch
JJ
,
et
al
.
P.
A
Co
m
par
ison
of
Sound
Segre
gation
Te
chn
iq
ues
for
Predom
i
nant
Instrum
ent
Rec
ognition
in
Mus
ic
al
Audio
Signal
s.
IRMA
S: a
da
ta
set
for
inst
rum
ent
re
cogni
t
i
on
in
m
usical
au
dio
signa
ls.
Evaluation Warning : The document was created with Spire.PDF for Python.