Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
3
,
J
un
e
201
9
, pp.
1871
~
18
78
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
9
i
3
.
pp1871
-
18
78
1871
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Modifie
d honey
encrypti
on sche
me
f
or
en
co
din
g natur
al
languag
e mess
age
Ab
io
dun
Esth
er Om
ola
r
a
,
Aman
Jantan
School
of Com
pu
te
r
Scie
nces,
Unive
rsiti
Sains
Ma
la
ysi
a, Pe
nang, M
al
ay
sia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
May
24, 201
8
Re
vised N
ov 27, 2
018
Accepte
d Dec
22, 201
8
C
onvent
ional
en
cr
y
p
ti
on
sch
emes
are
sus
ce
pti
b
le
to
brute
-
for
ce
a
t
ta
cks.
Thi
s
is
bec
ause
b
y
tes
enc
ode
utf8
(or
AS
CII)
cha
rac
t
ers.
Conseq
uent
l
y
,
an
adve
rsar
y
tha
t
i
nte
rc
ept
s
a
ci
ph
ert
ex
t
and
tries
to
dec
r
y
p
t
th
e
m
essage
b
y
brute
-
f
orc
ing
wit
h
an
in
cor
re
ct
k
e
y
ca
n
f
il
t
er
out
som
e
of
the
combinat
ions
o
f
the
d
ec
r
y
pt
ed
m
essage
b
y
ob
servi
ng
th
at
so
m
e
of
the
sequ
enc
es
are
a
combinat
ion
o
f
cha
ra
cters
whic
h
are
d
istri
but
ed
non
-
unifor
m
l
y
and
form
no
pla
usible
m
ea
ni
ng.
Hone
y
en
cr
y
pt
ion
(HE)
sch
eme
was
proposed
to
cu
rtail
thi
s
vulne
rab
il
i
t
y
of
conve
n
ti
o
nal
enc
r
y
p
ti
on
b
y
produc
ing
ci
pher
t
ext
s
y
i
e
ldi
ng
va
li
d
-
l
ooking,
unifor
m
l
y
distri
but
ed
but
fak
e
pla
i
nte
xts
upon
dec
r
y
pt
ion
with
inc
orre
ct
k
e
y
s
.
How
eve
r,
the
sche
m
e
works
for
on
l
y
passw
ords
and
PIN
S.
It
s
ada
pt
at
ion
to
support
enc
od
ing
nat
ur
al
l
angua
g
e
m
essage
s
(
e
-
m
ai
ls,
hum
an
-
gen
e
rat
ed
do
cument
s)
ha
s
remai
ne
d
an
open
proble
m
.
Exi
sti
ng
proposal
s
to
ext
end
th
e
sche
m
e
to
support
enc
odin
g
nat
ura
l
la
ngu
ag
e
m
essage
s
rev
ea
ls
fra
gm
ent
s
of
the
plaint
ext
in
th
e
ci
pher
te
x
t
,
hen
c
e,
it
s
sus
ce
pt
ibi
l
ity
to
chose
n
ciphert
ex
t
attac
ks
(CCA)
.
In
thi
s
pape
r,
we
m
odif
y
th
e
HE
sche
m
es
to
sup
por
t
the
enc
od
in
g
of
nat
ura
l
la
nguag
e
m
essage
s
using
Natur
a
l
La
ngu
age
Proc
essing
te
chn
iques
.
Our
m
ai
n
cont
ribution
was
creat
ing
a
struc
ture
tha
t
allowe
d
a
m
essage
to
be
e
n
code
d
ent
ir
ely
in
bina
r
y
.
As
a
result
of
thi
s
strat
eg
y
,
m
ost
bina
r
y
stri
ng
produc
es
s
y
ntactica
lly
cor
rec
t
m
essage
s
which
wi
l
l
be
gen
era
t
ed
to
decei
v
e
an
a
ttacke
r
who
at
te
m
pts
to
dec
r
y
pt
a
ci
ph
e
rte
xt
using
inc
or
rec
t
k
e
y
s.
W
e
e
val
ua
te
th
e
sec
urity
of
our
p
roposed
sche
m
e.
Ke
yw
or
d
s
:
Adve
rsar
y
Atta
cker
Brute
-
force
Ci
ph
erte
xt
Honey e
ncr
y
ption
P
l
ai
ntext
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
:
A
m
an
Jantan
,
Secu
rity
an
d F
or
e
ns
ic
Resea
r
ch Gr
oup (S
FRG)
Lab
orat
or
y
,
School
of Com
pu
te
r
Scie
nces,
Un
i
ver
sit
i
Sain
s Mal
ay
sia
, Mal
ay
sia
.
Em
a
il
:
sty
le
es
t2011@
gm
ail.co
m
1.
INTROD
U
CTION
Crypto
gr
a
phy
li
es
at
the
hear
t
of
a
m
ulti
tud
e
of
pr
ocesses
t
hat
inco
r
porate
E
-
Com
m
erce,
In
te
r
net
Ba
nki
ng,
Busi
ness
and
So
ci
al
Ne
tworks
,
Sm
art
Syst
e
m
s
and
so
f
ort
h
in
our
i
nfor
m
at
ion
-
awa
re
so
ci
et
y
[1
]
,
[
2]
.
T
he
e
vo
l
utio
n
of
the
d
igit
a
l
world
has
af
f
ect
ed
the
way
in
w
hich
pe
op
le
ma
nag
e
fina
nces,
proc
ur
e
goods
and
acce
ss
he
al
thcare.
Pe
op
l
e
hav
e
beco
m
e
increasin
gly
r
el
ia
nt
on
le
ve
r
agin
g
te
ch
no
l
ogy
to
enjoy
instant
acce
ss
to
in
for
m
at
ion
,
bu
si
ne
ss
colla
borati
on,
fam
il
y
and
fr
ie
nds
[3
]
,
[
4]
.
This
trend
of
so
ci
et
al
dep
e
ndence
on
m
od
er
n
in
f
rastr
uctur
e
ha
s
create
d
a
n
avenue
for
c
on
sta
nt
sec
ur
i
ty
threats
an
d
ris
k.
This
chall
en
ge
has
bee
n
dem
on
st
rated
by
in
cessant
at
ta
cks
on
c
om
pu
te
r
n
et
work
s
,
cre
dit
card
fr
a
ud,
m
a
lware
at
ta
cks,
cel
l
ph
on
e
s
hac
k
et
c
et
era
[
5],
[
6]
.
To
reinfo
rce
our
c
om
pu
te
rize
d
-
reli
ant
s
ociet
y
for
the
prese
nt
an
d
fu
t
ur
e
gen
e
rati
on,
an
d
le
ve
ra
ge
on
the
op
portu
niti
es
that
t
hese
te
ch
no
l
ogie
s
proffer,
t
he
re
is
a
pr
essi
ng
ne
e
d
for
reli
able
in
form
ation
in
frast
ru
ct
ure
with
str
ong
sec
uri
ty
req
uirem
ents.
A
n
esse
ntial
bu
il
di
ng
bl
ock
t
o
achieve
i
nfo
rm
at
ion
sec
ur
it
y i
s cr
ypt
ogra
ph
y
[
7], [8]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
18
7
1
-
1878
1872
Crypto
gr
a
phy
is
the
stud
y
of
the
sci
ence
and
a
rt
of
c
on
c
eal
ing
inf
orm
a
ti
on
to
pre
ven
t
a
distru
ste
d
par
ty
f
ro
m
le
a
rn
i
ng
t
he
co
nt
ent
of
the
m
essage
[
9]
.
Stat
e
-
of
-
t
he
-
a
rt
c
on
ven
ti
onal
cry
pto
gra
ph
ic
sc
he
m
es
are
pro
ne
to
brute
-
force
at
ta
ck.
A
brute
force
at
ta
ck
is
a
t
rial
and
er
ror
a
pproach
us
ed
by
a
n
ad
ver
sa
ry
to
de
cryp
t
a
ci
phertex
t,
f
or
insta
nc
e,
an
a
tt
e
m
pt
to
crac
k
a
us
er
’s
pass
w
ord.
T
he
c
onve
ntion
al
enc
ryption
sc
hem
es
m
ake
it
easy
for
a
n
a
dv
e
rsa
ry
to
c
onti
nu
e
to
em
pl
oy
this
m
et
ho
d
an
d
ac
hieve
s
uccess
.
I
n
hind
sigh
t,
a
ll
co
nv
e
ntion
a
l
encr
y
ption
sch
e
m
es
are
enc
ode
d
in
utf
8
(or
AS
C
II)
c
har
a
ct
ers
.
C
on
se
qu
ently
,
an
a
dver
sary
that
inte
rc
epts
a
ci
ph
e
rtext
a
nd
trie
s
to
de
crypt
the
m
essage
usi
ng
a
w
ron
g
ke
y
can
filt
er
out
so
m
e
of
t
he
com
bin
at
ion
s
of
t
he
decr
y
pted
m
essage
by
obse
r
vi
ng
t
hat
s
om
e
of
t
he
se
quenc
es
are
a
com
bin
at
ion
of
c
harac
te
rs
f
ro
m
dif
fer
e
nt
la
nguag
e
s
an
d
al
so
sym
bo
ls
wh
ic
h
are
no
n
-
unif
or
m
ly
distribu
te
d
[
10]
,
[
11
]
.
It
is
safe
to
co
nclu
de
th
at
an
adv
e
rsa
ry
can
j
ud
ge
if
his
at
ta
ck
is
su
ccessf
ul
based
on
th
e
structure/
distrib
ution
of
the
m
essage
he
re
cov
e
rs
durin
g
his
at
ta
cks
[
1
2
]
,
[
13
]
.
A
non
-
unifo
r
m
distribu
ti
on,
wh
ic
h
f
or
m
s
no
plausib
le
m
eaning
sig
nif
ie
s
an
inco
rr
ect
plai
nt
ext
and
key
.
More
ov
e
r,
t
he
adv
e
nt
of
hi
gh
-
proce
ssin
g
to
ols
su
c
h
as
F
P
GA,
G
PU
al
s
o
m
ake
s
the
bru
t
e
-
force
at
ta
ck
to
be
highly
su
cce
s
s
fu
l
[
14]
,
[15],
[16].
The
refo
re,
a
n
at
ta
cke
r
that
intercept
s
an
encr
y
pted
m
essage
has
a
hi
gh
c
han
ce
of
re
cov
e
rin
g
t
he
ke
y
and
s
ub
se
quently
the
m
essage
e
ncr
y
pted
unde
r
su
c
h
key
by
us
in
g
the
disti
nguis
her
te
ch
ni
qu
e
–
w
hich
is
to
judge
the
decr
y
ption
ou
t
pu
t
base
d
on
t
he
distrib
ution
/st
r
uctu
re
of
t
he
m
essage
.
S
ecu
rity
exp
er
ts
and
the
c
r
yptogra
ph
ic
c
omm
un
it
y
hav
e
propose
d
se
ver
al
c
ounter
m
easur
es
t
o
m
itigate
the
brute
-
f
or
ce
at
ta
ck
on
c
onve
nt
ion
al
e
ncr
ypti
on
schem
es
by
increasin
g
t
he
le
ngth
of
the
key,
com
pu
ta
ti
on
al
com
plex
ity
.
N
ever
t
heless,
al
l
these
m
et
hods
fail
to
ad
dress
the
unde
rly
in
g
pro
blem
,
whic
h
is
us
in
g
the
disti
nguis
her
te
ch
ni
qu
e
to
i
nfor
m
the
outp
ut
of
the
dec
ryptio
n.
A
n
at
ta
cke
r
c
an
co
ntin
ue
to
m
ake
rep
eat
dec
ryption
s
on a
n o
racle by try
in
g ran
do
m
k
ey
s
un
ti
l he
fin
ds
a
pla
usi
b
le
m
essage
[1
7
-
18]
.
A
br
ea
kthro
ug
h
was
m
ade
in
the
ye
ar
2014
w
hen
Ju
el
s
a
nd
Ri
ste
np
art
[
19
-
20]
pro
posed
the
H
oney
Encr
y
ption
(
H
E)
sc
hem
e
in
a
co
nf
e
ren
ce
,
E
ur
oc
rypt
2014
.
H
on
ey
e
ncr
y
pt
ion
is
a
par
a
di
gm
that
pr
ovid
es
an
d
i
m
pr
oves
sec
ur
it
y
beyond
t
he
conve
ntion
al
brute
-
f
or
ce
bo
unds.
It
prov
i
de
s
a
plausi
ble,
va
li
d
-
lo
ok
i
ng
but
fa
ke
m
essage when
an
inc
orrect ke
y
is used
f
or d
e
crypti
ng a cip
he
rtext. I
t
do
e
s
no
t y
ie
ld un
der a b
r
ute
-
force a
tt
ack
and
it
was
pro
po
s
ed
f
or
pass
words
a
nd
PIN
s
an
d
hav
e
bee
n
im
ple
m
ented
in
t
his
reg
a
r
d
su
ccess
fu
ll
y
[
19
-
20]
.
Howe
ver,
HE
has
no
t
bee
n
ap
plied
for
the
secu
rity
of
natu
ral
la
ngua
ge
m
essage
s
s
uch
as
e
m
ai
ls,
hu
m
an
-
gen
e
rat
ed
doc
um
en
t
s
and
ce
rtai
n
f
orm
at
s
of
m
essage
.
T
he
ap
plica
ti
on
s
of
the
st
and
a
r
d
HE
pro
po
s
ed
by
[
1
9
-
20
]
a
r
e
preci
sel
y
for
p
ass
word
s
t
o
protect
pa
ss
words
or
lo
w
m
in
-
ent
ropy
keys
to
protec
t
high
m
in
-
entr
op
y
ke
ys.
I
nd
ee
d,
th
e
auth
ors
of
H
E
le
ft
as
a
n
open
chall
e
ng
e
,
it
s
ada
ptati
on
to
natu
ral
la
ngua
ge
m
essage wh
e
n t
hey m
ade th
is st
at
e
m
ent,
“
...
fo
r
hu
man
-
gen
e
ra
te
d
m
es
sa
ge
s
(
p
as
sw
ord
v
au
lt
s,
e
-
m
ail,
et
c.
)
(
...
)
i
s
intere
sti
ng
as
a
natur
al
langua
ge pr
oc
essing pr
oble
m
.
” [
19
]
, [2
0
]
.
Wh
il
e
so
m
e
researc
h
ef
f
or
ts
ha
ve
be
en
m
ade
towards
a
dap
ti
ng
the
sc
hem
e
to
enc
od
e
hu
m
an
-
gen
e
rat
ed
m
essage
,
the
current
proposals
are
al
l
s
us
ce
ptible
t
o
a
Cho
se
n
-
Ci
ph
e
rtext
Atta
ck
(
CC
A)
.
A
CCA
is
a
n
at
ta
ck
m
od
el
fo
r
c
ryptanaly
si
s
in
w
hich
t
he
adv
e
rsa
ry
col
le
ct
s
inform
ati
on,
at
le
ast
in
par
t,
by
sel
ect
ing
a
ci
ph
e
rtext
an
d
decr
y
pting
it
by
try
ing
rand
om
keys.
This
at
ta
ck
is
inev
i
ta
ble
wh
e
n
pa
rts
or
fr
a
gm
en
ts
of
t
he
plai
ntext
m
essage
is
rev
ea
le
d
in
the
dec
r
ypte
d
m
essage.
A
pr
act
ic
al
a
t
ta
cker
ca
n
easi
ly
us
e
the
re
vealed
pa
rts
of
the
m
essage
to
f
or
m
oth
er
par
ts
of
t
he
m
essage
an
d
recove
r
the
plainte
xt
an
d
the
key
.
More
ov
e
r
,
s
ome
of
t
he
pr
opose
d
ap
proac
h
es
do
n
ot
scal
e
w
el
l;
they
fail
to
m
od
el
hum
an
la
nguag
e
w
he
n
us
e
d
to en
c
ode
h
u
m
an
-
ge
ner
at
e
d d
ocu
m
ents, thus
, f
ai
li
ng to
con
vin
ce t
he
at
ta
c
ker that
he has
the tar
get m
ess
age.
This
pa
per
see
ks
to
ad
dr
ess
t
he
shortfall
s
of
the
sta
nd
a
rd
Honey
encr
y
ption
schem
e
and
pro
posal
s
m
ade
by
the
nota
ble
res
earch
ef
forts
in
this
dom
ai
n
.
T
he
pr
i
ncipal
obj
ect
ive
is
t
o
e
xtend
the H
E schem
es
to
ada
pt
i
t
for
encodin
g
natu
r
al
la
nguag
e
m
essage.
O
ur
a
ppr
oac
h
al
so
a
ddres
se
s
the
C
CA
at
ta
cks
whic
h
cu
rr
e
nt
ap
proac
h
fail
to
cur
ta
il
.
We
em
plo
ye
d
Natu
ral
la
ngua
ge
proce
ssin
g
too
ls
an
d
te
ch
niques
to
buil
d
co
nv
i
ncin
g
deco
y
m
essage
wh
ic
h
will
be
ser
ved
to
a
n
at
ta
cker
that
trie
s
to
decr
ypt
a
ci
ph
erte
xt
usi
ng
inc
orrect
keys
.
This
strat
egy
will
ultim
a
te
ly
al
low
encr
y
pt
ion
of
oth
e
r
form
at
s
sp
eci
fical
ly
hu
m
an
-
gen
e
rated
m
essages,
s
u
ch
as
em
ai
ls
,
w
ritt
en
doc
um
ents,
et
c.
Fi
gure
1
prese
nts
how
t
he
c
onven
ti
onal
e
ncry
ption
sc
hem
e
wor
ks
durin
g
a
bru
te
-
force att
ack
.
I
n sum
m
a
ry
,
the co
nt
rib
ution
s
of this
resea
rch
inclu
de
the
foll
ow
i
ng
:
a.
C
reati
ng
a
str
uc
ture
t
hat
al
lo
ws
a
m
essage
to
be
e
ncode
d
entir
el
y
in
bina
ry.
As
a
res
ul
t
of
the
e
nc
odin
g
strat
egy,
m
os
t
bin
ary
stri
ng
s
produce
a
sy
ntact
ic
al
ly
cor
rect
m
essage
wh
ic
h
is
capa
ble
of
foolin
g
an
adv
e
rsa
ry in se
arch o
f
the
p
la
i
ntext a
nd k
ey
.
b.
The
pro
pose
d
schem
e
add
res
s
CC
A
at
ta
ck
by
gen
erati
ng
a
m
essage
struct
ur
e
that
does
not
rev
eal
any
pa
r
t
of
the
plainte
xt
in
the
deco
y
m
essage
and
f
ai
le
d
decr
ypti
on
(
inc
orrect
ke
y)
produces
ra
dical
ly
diff
ere
nt
m
essages fr
om t
he plai
ntext.
c.
Len
gth
of
plainte
xt
is
co
ncea
le
d
by
ap
pe
nd
i
ng
a
rand
om
bin
ary
stri
ng
to
the
ci
pherte
xt
s
ince
the
d
eco
de
r
runs
unti
l
the
sentence
is
pars
ed
an
d
then
sto
ps
.
T
his
way
the
at
ta
cker
is
unable
to
dete
r
m
ine
the
le
n
gth
of
the p
la
inte
xt
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Mo
difi
ed
ho
ne
y encry
ptio
n
sc
hem
e
for
en
c
odin
g na
t
ural l
angu
ag
e
m
e
ssag
e
(
Abiodun
Esther
Omola
ra
)
1873
d.
First
su
ccess
f
ul
i
m
ple
m
entat
i
on
of
a
pro
va
bl
e
secur
e
d
nat
ur
al
-
la
ngua
ge
-
base
d
hone
y
encr
y
ption
sc
he
m
e
and it
s
ada
ptati
on to natu
ral la
ngua
ge
m
essages.
Figure
1
.
Br
ute
-
f
or
ce
att
ack
on
c
urren
t
c
onve
ntion
al
enc
ryption m
et
ho
d
on a
pass
w
ord v
ault
An
at
ta
cker
th
at
intercepts
a
plainte
xt
a
nd
s
ub
m
it
s
rando
m
pass
w
ords
get
non
-
un
if
orm
l
y
distrib
ute
d
char
act
e
rs
as
lon
g
as
he
sup
plies
an
inco
rrec
t
key.
This
is
a
cl
ear
disti
ng
uis
her
that
he
has
no
t
achie
ved
the
plainte
xt.
He
con
ti
nues
t
o
try
rand
om
keys
un
ti
l
he
ge
ts
a
valid
-
lo
okin
g
m
essage.
As
soo
n
as
he
gets
plausi
ble,
un
i
f
or
m
ly
distribut
ed
c
har
act
ers
,
then
he
is
s
ur
e
he
ha
s
ac
hieve
d
t
he
m
e
ssage
a
nd
t
he
key
.
Fo
r
exam
ple,
in
Fig
ure
1,
w
hen
t
he
at
ta
ck
er
gets
‘12
3456’
,
he
knows
t
hat
he
has
ac
hi
eved
t
he
key.
He
is
certai
n
he
ha
s
t
he
correct
key
becau
se
‘12
3456’
is
distrib
ut
ed
eve
nly
and
this
is
m
ade
wo
rst
as
us
e
rs
te
nd
to
us
e
sim
ple,
e
as
y
-
to
-
guess
pa
sswor
ds
.
Als
o,
the
a
t
ta
cker
s
are
awa
re
of
how
use
rs
c
hoos
e
wea
k
pa
ss
words
(b
ase
d o
n
the
past
r
el
ease
of
c
om
pr
om
ise
d
p
a
sswor
ds
).
2.
HON
E
Y
E
NCRY
PTI
ON
The
int
uiti
on
of
Honey
e
ncry
ption
was
born
e
by
Ju
el
s
a
nd
Ri
ste
npa
rt
[19
-
20]
after
obser
ving
th
e
high
rate
of
pa
sswor
d
be
e
n
c
om
pr
om
ise
d
.
F
or
instance
,
usi
ng
c
onve
ntio
na
l
encr
y
ption
s
chem
e,
if
plain
te
xt,
P
i
s
a
16
-
di
git
Ma
ste
r
card
nu
m
ber
enc
od
ed
via
A
SCI
I
and
t
he
co
nv
e
ntion
al
pass
w
ord
-
base
d
enc
r
ypti
on
schem
e
is
us
ed
as
c
iph
e
r,
t
he
pro
bab
il
it
y
that
any
Pi≠
P
is
a v
al
id A
SCI
I
en
cod
i
ng
of
a 1
6
-
dig
it
strin
g
wh
i
ch
is
neg
li
gi
ble,
at
t
(
10
/
256)
16
<
2
−74
.
He
nce
,
an
a
dversa
ry
c
an
rej
ect
inc
orrect
m
essages
and
rec
over
P
with
a
high
pro
bab
il
it
y
[19
-
20]
.
Honey
enc
rypt
ion
prov
i
des
s
ecur
it
y
beyo
nd
the
brute
-
f
or
c
e
barrier
by
producin
g
ci
phe
r
te
xts
wh
ic
h,
upon
decr
y
ptio
n
with
inc
orrec
t
keys,
yi
el
ds
pl
ausible
-
lo
okin
g
plainte
xts.
T
he
strat
e
gy
he
r
e
is
to
c
onf
ound
the
li
fe
of
the
a
dv
ersar
y
by
m
aking
m
essage
re
cov
e
ry
eve
n
af
te
r
try
ing
e
very
cand
idate
ke
ys
to
be
im
po
ssible.
The
ce
ntral
c
om
po
ne
nt
of
HE
is
cal
le
d
a
distrib
utio
n
trans
form
ing
encode
r
(
DTE
)
.
This
c
om
po
ne
nt
is
const
ru
ct
e
d
wi
th
an
est
im
at
e
of
the
m
essage
distrib
ution
pm
.
Encodin
g
a
m
essage
sam
pled
f
r
om
p
m
sh
oul
d
pro
du
ce
a
“s
eed”
val
ue
wh
ic
h
is
uni
form
l
y
distribu
te
d
.
The
D
TE
is
a
pai
r
of
enc
odin
g
an
d
decodin
g
al
gor
it
h
m
.
The
e
ncodin
g
al
gorithm
is
ra
ndom
iz
ed,
it
tak
es
as
in
put
a
m
essage
M
∈
M
and
outp
uts
a
value
i
n
a
set
S.
Also,
the
dec
od
i
ng
al
go
rithm
is
determ
inist
ic
and
ta
ke
s
as
in
put
a v
al
ue
S
∈
S
an
d
outp
uts
a
m
essage
M
∈
M.
Wh
e
n
th
e
seed
is
sup
plied,
a
DTE
m
ust
decode
c
orr
ect
ly
to
the
pl
ai
ntext
but
f
or
ever
y
i
ncorr
e
ct
key,
it
sh
ould
decode
t
o
a
pla
us
ible
m
essage
s
a
m
pled
f
r
om
t
he
m
essage
distribu
ti
on.
A
DTE
schem
e
is
co
rr
ec
t
if for
an
y
M
∈
M
,
Pr[de
co
de(enco
de
(M)) =
M]
= 1
[19
-
20]
.
The
HE
sc
he
m
e
con
sist
s
of
two
ste
ps
usu
al
ly
ref
err
e
d
t
o
as
DTE
-
then
Enc
rypt
.
I
n
the
fi
rst
ste
p
,
the
DTE
is
ap
pl
ie
d
to
the
Me
s
sage,
M
to
obt
ai
n
a
See
d
S.
I
n
the
sec
ond
ste
p,
the
seed
S
is
encr
ypte
d
unde
r
a
conve
ntion
al
e
ncr
y
ption sche
m
e u
sin
g
the
key
K
, yie
ldin
g an
HE
ci
ph
e
rte
xt C.
T
he st
eps
are
as
foll
ows:
a.
The
enc
odin
g
process
,
w
her
e
the
Distrib
ution
T
ra
ns
f
or
m
i
ng
E
nc
o
de
r
(
DTE
)
is
app
li
ed
to
the
plain
te
xt
fo
ll
owe
d by a
conve
ntion
al
e
ncr
y
ption sche
m
e.
b.
The
dec
od
i
ng
process
,
wh
e
re
the
Distrib
ution
Tra
nsfo
rm
i
ng
Dec
oder
(
D
TD)
is
a
pp
li
ed
to
th
e
ci
phe
rtext
fo
ll
owe
d by a
conve
ntion
al
e
ncr
y
ption sche
m
e.
The HE al
gorithm
sh
ow
i
ng
th
e enc
od
e
and
de
cod
e
pr
ocess
i
s prese
nted
i
n Fi
gure
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
18
7
1
-
1878
1874
H
o
ney
Encry
ptio
n Algo
rith
m
:
HEnc (
K,
M)
S ¬$ en
co
d
e(M
)
R ¬$
{0
,
1
}n
S’
¬
H (
R
,
K
)
C ¬ S’
⊕
S
return (R,
C)
H
o
ney
Dec
ry
ption Alg
o
rith
m
:
HDec
(K,
(
R,
C))
S’¬
H
(
R,
K)
S ¬
C
⊕
S
’
M
¬ deco
d
e(S
)
return M
Figure
2. H
one
y enco
ding a
nd
d
ec
od
i
ng
al
go
rithm
s
H
re
pr
ese
nts
the
crypt
ogra
phic
has
h
f
un
c
ti
on
,
K
re
pr
es
ents
a
key
,
M
re
presents
th
e
m
essage
,
S
re
pr
ese
nts
th
e
seed,
R
re
presents
a
rand
om
s
tring
,
C
re
pr
ese
nts
the
ci
ph
e
rtext,
a
nd
¬$
re
presents
un
i
form
rand
om
assignm
ent.
3.
PROP
OSE
D
APP
ROAC
H
Our
ap
proac
h
le
ver
a
ges
the
powe
r
of
dece
pt
ion
to
per
s
ua
de
and
c
onfuse
the
at
ta
cker
th
at
he
has
the
or
i
gin
al
m
essage.
The
ex
per
i
m
ent
of
this
re
search
are
im
p
lem
ented
with
Pyt
hon
3.63
a
nd
Nat
ur
al
La
ngua
ge
Pr
oc
essin
g
(NLP)
li
br
a
ries.
The
NL
P
to
ols
capt
ur
e
the
rel
at
ion
s
hip
betw
een
w
ords
a
nd
how
they
com
bin
e
to
form
m
eaningf
ul
sente
nces.
Re
ader
s
i
ntere
ste
d
in
detai
l
t
reatm
ent
of
N
LP
sho
uld
see
[21]
-
[
22]
.
Th
e
decoy
m
essage
will
hav
e
a
c
om
ple
te
l
y
diff
eren
t
m
essage
struct
ur
e
f
r
om
the
or
iginal
m
essage
(p
la
intext
)
but
will
hav
e
a
sem
antic
and
co
nte
xtua
l
m
eaning
.
Ea
ch
m
essage
to
be
enc
od
e
d
is
treat
ed
as
a
string
of
sente
nc
e
an
d
processe
d
as
a
gr
am
m
atical
featur
e
in
t
he
En
glish
La
ngua
ge.
Eac
h
wor
d
is
cl
assifi
ed
base
d
on
the
pa
rt
of
sp
eec
h
by
it
s
s
ynta
ct
ic
functi
on
s
.
In
the
E
ngli
sh
La
ngua
ge,
the
m
ai
n
pa
rts
of
s
peec
h
a
re
noun,
pro
noun,
ve
rb,
adv
e
r
b,
ad
j
ect
ive,
prep
os
it
io
n,
inter
j
ect
io
n,
conjuncti
on
and
determ
iner.
Fig
ur
e
3
s
hows
how
nouns
are
processe
d
a
nd
encr
y
pted,
an
d
Fig
ur
e
4
s
ho
ws
a
flo
wch
a
rt
of
the
c
om
plete
syst
e
m
of
our
pro
posed
sc
hem
e
.
Figure
5
s
how
s
how
nouns
a
re
pro
cesse
d
a
nd
dec
rypted
and
Fig
ur
e
6
s
hows
a
flo
wc
ha
rt
of
the
c
omplet
e
syst
e
m
of
our
pro
po
se
d
sche
m
e.
In
s
umm
a
ry,
we
c
reate
a
struc
ture
tha
t
al
lowe
d
a
se
ntence
to
be
e
ncode
d
com
plete
ly
in
bin
ary.
As
a
resu
lt
of
the
encodin
g
stra
te
gy,
m
os
t
bi
nar
y
strin
gs
pro
duce
a
synt
act
ic
al
ly
correct
se
ntenc
e.
Alg
o
rith
m
1: No
u
n
E
n
co
d
in
g
1
: Get
syns
ets f
o
r
t
h
e
noun
if
perso
n
al no
u
n
Drop
sy
n
sets
with
n
o
n
-
p
erso
n
al
m
ean
in
g
f
o
r
th
e
n
o
u
n
(
i.e.
,
u
sin
g
m
u
m
to r
ef
er
to
a
f
lo
wer)
Else
Drop
sy
n
sets
with
p
erso
n
al
m
eani
n
g
f
o
r
th
e
noun
(
i.e.
,
u
sin
g
a
dog
to
r
ef
er
to
a
p
erso
n
)
if
sy
n
set list is n
o
n
-
e
m
p
t
y
Ran
d
o
m
l
y
select
s
y
n
set f
ro
m
list
Else
if
th
e
noun
is
a
p
ro
n
o
u
n
Set of
f
set to
the len
g
th
o
f
th
e
list o
f
n
o
u
n
s in
word
n
et pl
u
s o
n
e and
enco
d
e
in
bin
ary
Enco
d
e pro
n
o
u
n
Else
Set of
f
set to
the len
g
th
o
f
th
e
list o
f
n
o
u
n
s in
word
n
et pl
u
s
two
and
enco
d
e in b
in
ary
Enco
d
e un
k
n
o
wn
word
return
b
in
ary strin
g
2
: Get
th
e
p
ath
f
rom
th
e
roo
t
o
f
word
n
et no
u
n
tr
ee to selected
sy
n
set
if
perso
n
al no
u
n
Drop
no
d
es in
the p
ath
that are
p
ar
en
ts to
"
p
erso
n
no
u
n
"
syns
et
Else
Drop
no
d
es in
th
e
p
ath
u
n
til "
p
erso
n
no
u
n
"
syns
et is no
t in
th
e
p
ath
Ran
d
o
m
l
y
select
a
n
o
d
e in
th
e
re
m
ain
in
g
p
ath
to b
e su
b
t
ree
roo
t
Enco
d
e of
fset of
th
e
selected
n
o
d
e in
t
h
e
list
o
f
word
n
et's n
o
u
n
s in
bin
ary
wh
ile no
t at
th
e
d
e
sired
n
o
d
e (
n
o
u
n
p
ass
ed
in)
Ap
p
en
d
1 to
bin
ar
y
enco
d
in
g
to in
d
i
cate that
se
a
rch
is
n
o
t co
m
p
lete
Fin
d
th
e
in
d
ex
o
f
n
ex
t no
d
e in
th
e
p
ath
in
th
e
list
o
f
curr
en
t r
o
o
t's ch
ild
ren
Enco
d
e ind
ex
in b
i
n
ary
and
app
en
d
to
a
strin
g
Ap
p
en
d
0 to
bin
ar
y
enco
d
in
g
to in
d
i
cate that
se
a
rch
is
co
m
p
let
e
Fin
d
ind
ex
in
th
e
list
o
f
sy
n
set's
le
m
m
as o
f
le
m
m
a
con
tai
n
in
g
th
e
d
esired
n
o
u
n
Enco
d
e ind
ex
in b
i
n
ary
and
app
en
d
to
a
strin
g
3
: r
etu
rn b
in
ary str
in
g
Figure
3
.
Pse
udoc
ode
for
im
ple
m
enting
Enc
rypt
Algorithm
of the
Pro
pose
d (Mo
dified
) H
on
ey
Enc
ryptio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Mo
difi
ed
ho
ne
y encry
ptio
n
sc
hem
e
for
en
c
odin
g na
t
ural l
angu
ag
e
m
e
ssag
e
(
Abiodun
Esther
Omola
ra
)
1875
Figure
4
.
Pro
pose
d
e
ncodin
g al
gorithm
Alg
o
rith
m
2: No
u
n
Decod
in
g
1
: E
x
tract
o
f
f
set f
ro
m
bin
ary strin
g
if
pro
n
o
u
n
return d
ecod
in
g
of
th
e
p
ron
o
u
n
if
un
k
n
o
wn
word
return d
ecod
in
g
of
th
e
u
n
k
n
o
w
n
wo
rd
2
: Get
syns
et
of
a
noun
at giv
en
o
f
fs
et in
th
e
list
o
f
wor
d
n
et's n
o
u
n
s an
d
s
et as r
o
o
t
wh
ile no
t at
th
e de
sired
no
d
e (
th
e f
irst characte
r
in
th
e
b
in
ary
st
ring
is no
n
-
zero
)
Extract of
f
set f
ro
m
bin
ary
st
ring
Set child
at
o
f
f
set in
the list o
f
r
o
o
t's ch
ild
ren to
new roo
t
3
: E
x
tract
l
e
m
m
a
n
u
m
b
e
r
f
ro
m
a
b
in
a
ry
st
ring
4
: Get
sp
ecif
ied
le
m
m
a
5
: r
etu
rn
th
e
n
a
m
e
o
f
th
e
le
m
m
a
(des
i
red n
o
u
n
)
Figure
5
.
Pse
udoc
ode
for
im
ple
m
enting
de
cr
ypt
al
gorithm
of the
prop
os
e
d (m
od
ifie
d) ho
ne
y encry
ption
Figure
6
.
Pro
pose
d
decodin
g al
gorithm
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
18
7
1
-
1878
1876
Note
that
:
Nouns
in
Wor
dn
et
a
re
orga
ni
zed
usi
ng
a
tre
e
structu
re
with
ge
ne
ral
co
nc
epts
at
the
r
oo
t
,
m
ov
ing
to
spe
ci
fic
nodes
at the le
aves.
W
e
u
ti
li
ze this str
uctu
re
for
e
ncodin
g o
f
al
l n
ouns wit
hin
t
he
e
ncodi
ng alg
or
it
hm
.
Verb
le
m
m
as
in
Wo
r
dnet
are
store
d
with
li
sts
of
v
er
b
fr
am
es.
The
se
are
t
he
possi
ble
str
uctu
res
of
se
nt
ences
us
in
g
t
hat
ve
r
b.
F
or
e
xam
ple,
"So
m
ebody
ap
pear
s"
is
a
valid
sente
nc
e
struct
ur
e
but
"So
m
ebody
a
pp
ea
rs
so
m
et
hin
g"
is
no
t.
Pron
oun enc
odi
ng
/dec
od
in
g
: Hav
e
a list
of
pro
nouns,
e
nc
oded
as
a
n
offse
t
in
the
li
st
.
Un
k
no
w
n
enc
od
in
g/deco
din
g:
Each
let
te
r
e
ncode
d
as
a
posit
ion
in
the al
ph
a
bet
(0
-
25)
Art
ic
le
enc
od
i
ng
/dec
od
in
g
:
0/1
f
or
not exis
ts/
exists, the
n 1/0
i
f
the/a
Conjuncti
on
e
ncod
in
g/dec
odi
ng
:
Hav
e
a list
o
f
con
j
unct
i
on
s
, e
nc
od
e
d
a
s an o
ff
set
i
n
t
he
li
st
.
A
gen
e
ral
us
e
case
of
the
pro
po
se
d
al
gorithm
is
to
co
m
bin
e
it
with
any
c
onve
ntion
al
ci
phe
r
.
Fo
r
i
ns
ta
nce,
t
he
sen
der
will
encr
y
pt
this
bin
ary
strin
g
wit
h
a
ci
ph
e
r
(
AE
S,
RC
4,
T
wofish,
et
c.
)
.
A
wrong
key
will
produce
an
inc
orrect
ly
encode
d
stri
ng
w
hich
s
hould
decode
to
a
se
m
antic
ally
cor
rect
m
e
ssage
.
On
ly
wh
e
n
the
co
rr
ect
key
is
app
li
ed
s
hould
the
plainte
xt
de
cod
e
c
orrectl
y.
A
ny
inc
orre
ct
key
(
flip
ping
a
bit
)
will
decode
c
orrectl
y
to
a
val
id
m
essage
wi
th
com
plete
ly
diff
e
re
nt
le
ngth
a
nd
str
uctu
r
e
from
the
pla
intext.
Our
ap
proac
h
i
s
com
plete
ly
diff
e
ren
t
f
ro
m
existi
ng
ap
proac
h
beca
us
e
no
pa
rt
of
t
he
plain
te
xt
will
be
revea
le
d
durin
g
decr
y
ption
with
i
ncor
rect
key.
This
ap
proach
m
a
kes
it
im
po
ssible
f
or
the
a
dversa
ry
to
ca
rry
ou
t
a
chosen
ci
phert
ext att
ack.
We
c
on
si
der
a
case
w
he
re
an
em
a
il
is
to
be
enc
od
e
d
us
in
g
our
a
ppr
oac
h.
For
insta
nce
,
each
em
ail
us
ua
ll
y
sta
rts
with
so
m
e
gr
e
et
ing
an
d
will
us
ua
ll
y
end
wi
th
a
goodbye
m
essage.
An
a
dv
e
rsa
ry
exp
ec
t
ing
a
n
e
m
ai
l
w
ho
get
s
instea
d
a
te
xt
no
t
sta
rtin
g
f
r
om
this
ty
pe
of
gr
eet
in
gs
will
qu
ic
kly
discar
d
the
m
essage.
In
this
case
,
we
def
i
ne
encr
y
ption
prot
oco
l
diff
e
re
ntly
.
W
e
ass
um
e
that
e
m
a
il
-
m
essage
shou
l
d
al
ways
sta
rt
from
th
e
gr
eet
in
g
an
d
fi
rst
few
b
it
s
will
encode
one
of
the
possi
ble
greet
ing’s
va
riat
ion
s
.
This w
ay
it
is
no
t
po
ssi
ble
f
or
any
enc
ryptio
n
key
to
get
t
he
resu
lt
s
with
ou
t
greet
ing
s
an
d
goodbye.
T
herefo
re,
t
he
high
er
prob
a
bili
ty
of
t
he
deco
y
m
essage
is d
one
by r
est
rict
ing
pote
ntial
us
er
f
or
m
at
o
f
the
m
essage
.
An
eas
y
im
ple
m
entat
ion
wi
ll
be
to
de
fine
th
ree
pa
rts
of
t
he
m
es
sage
w
hic
h
a
re
al
lowe
d.
Fo
r
in
sta
nce,
1)
G
reeti
ngs
with
one
of
th
e
fo
ll
ow
i
ng
opti
ons:
00
-
"
Hell
o
",
01
-
"
Hi
",
10
-
"Gree
ti
ng
s
",
11
-
"G
ood
ti
m
e
of
day";
2)
m
idd
le
sentence
with
the
fo
l
lowing
opti
ons
:
00
-
"how
ar
e
you
?
";
01
-
"
Fine";
10
-
"h
avi
ng
a
ha
rd
day";
11
-
"l
et'
s
m
eet
"
;
3)
Goo
d
byes
:
00
-
"B
ye
";
10
-
"Lo
okin
g
f
orward";
01
-
"
ha
ve
a
nice
day";
11
-
"B
est
wis
hes
”
This
protoc
ol
will
lim
it
a
m
ou
nt
c
omm
un
ic
at
io
n
ver
y
m
uch
,
but
e
ver
y
com
bin
at
ion
or
key
can
le
ad
to
a
m
eaning
f
ul
m
essage.
T
her
e
is
no
way
fo
r
the
at
ta
ck
er
to
guess
if
his
key
was
i
ncorr
ect
.
4.
RESU
LT
S
AND A
N
ALYSIS
To
eval
ua
te
ou
r
pro
posed
sys
tem
,
we
te
st
with
a
con
tr
olled
am
ou
nt
of
noise
ad
ded
t
o
a
ci
ph
erte
xt
(
i.e.
,
1%
wr
ong,
5%
wron
g,
e
tc
.
)
a
nd co
m
par
e the
san
it
y
of r
es
ulti
ng
m
essages.
Wor
d
Co
unt:
T
he nu
m
ber
of
words e
ncode
d
No
ise
Added:
Perce
ntage of
ra
ndom
b
it
s f
li
pped
(
i
ncorr
ect
key)
Chan
ge
in
the
Wor
d
Count:
Wh
e
n
the
ad
ve
rsar
y
trie
s
to
de
crypt
with
the
key,
how
m
an
y
sane
wo
r
ds
wer
e
g
e
ner
at
e
d
.
P
im
por
tant
:
Ho
w
m
any
i
m
po
rtan
t
words
f
r
om
t
he
unde
rly
ing
plainte
xt
a
ppea
r
in
t
he
decr
y
pted
m
essage
wh
e
n
t
he
a
dversar
y
trie
s
in
correct
keys
.
We
d
efi
ne
P
im
portant
as
ver
y
im
po
rtant
key
words
in
the
plainte
xt
(no
un
s
,
pro
noun, ve
rb, a
dv
e
rb, adjecti
ve
) whi
ch
m
us
t no
t a
ppear
in
t
he
e
ve
nt of
decr
y
ptio
n wit
h wro
ng key
s.
P
regular
:
Ho
w
m
any
regular
words
f
ro
m
the
under
ly
in
g
pl
ai
ntext
app
e
ar
in
the
dec
ryp
te
d
m
essage
wh
e
n
t
he
ad
ve
rsar
y
trie
s
i
nc
orrect
keys.
We
def
in
e
P
reg
ular
as
com
m
on
w
ords
in
the
plainte
xt
(d
et
e
rm
iner,
conj
un
ct
io
n,
in
te
r
j
ect
io
n
,
pr
e
posit
ion).
T
his
is
essenti
al
bec
ause
re
gula
r
w
ords
help
f
orm
a
sentence
a
nd
they
app
ea
r
i
n
e
very
m
essage
a
nd
can
no
t
be
use
d
to
rec
over
the
plai
ntext.
Fo
r
e
xam
ple,
a,
the,
of
,
on
et
c
ar
e
essenti
al
to
f
orm
sentence
.
Ti
m
e
-
ta
ken
:
T
he
tim
e
i
t
ta
kes
to
decr
y
pt
us
i
ng
our
propose
d
m
et
ho
d.
Per
f
orm
ance
of the
pro
pose
d
m
od
ifie
d h
on
ey
en
cry
ption s
chem
e
as sh
ow
n
in
Ta
ble 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
Mo
difi
ed
ho
ne
y encry
ptio
n
sc
hem
e
for
en
c
odin
g na
t
ural l
angu
ag
e
m
e
ssag
e
(
Abiodun
Esther
Omola
ra
)
1877
Table
1.
Per
for
m
ance of th
e
P
rop
os
ed
Mo
dif
ie
d
H
oney
E
nc
ryptio
n
Sc
hem
e
W
o
rd Co
u
n
t
No
ise
Ad
d
ed
(%)
(bru
te
-
f
o
rcing
with
in
co
rr
ect ke
y
s)
Ch
an
g
e
in
th
e
wo
rd
an
d
wo
rd
co
u
n
t
d
u
ring
d
ecry
p
tio
n
Ch
eck
if
CC
A
i
s
p
o
ss
ib
le
(Any
revealed
wo
rd
f
ro
m
th
e
p
lain
tex
t
in
t
h
e decry
p
ted
ciph
ertext?
)
P
i
m
p
o
r
tan
t
P
r
eg
u
lar
Ti
m
e
-
tak
en
to
d
ecry
p
t
(
m
s
)
4
4
53
0
0
1
.90
7
2
59
0
4
2
.23
10
3
54
0
5
3
.12
15
2
62
0
4
3
.23
20
3
71
0
7
4
.40
25
4
88
0
6
4
.89
28
5
97
0
9
7
.10
38
1
105
0
6
9
.30
43
8
127
0
8
1
0
.11
48
7
156
0
9
1
3
.21
82
5
304
0
12
3
1
.11
5.
DISCU
SSI
ON
The
pro
pose
d
m
od
ifie
d
hone
y
encr
ypti
on
schem
e
e
m
plo
ys
a
no
vel
ap
pr
oach
for
ada
pting
the
honey
encr
y
ption
sch
e
m
e
to
sup
port
the
s
uccess
fu
l
enc
od
i
ng
of
hum
an
-
ge
ner
at
e
d
m
essage
w
hi
ch
has
bee
n
an
ope
n
pro
blem
since
the
sc
hem
e
was
pro
po
se
d
[19,2
0].
T
his
novel
i
nv
e
ntion
bri
ngs
ne
w
ben
e
fits
as
well
as
so
m
e lim
it
a
ti
on
s.
In
s
umm
ary,
we
we
re
able
to
create
a
structu
re
w
her
e
m
essages
are
encode
d
in
bi
nar
y
su
c
h
that
m
essage
rec
o
ve
ry
bec
om
es
im
po
ssible
f
or
an
a
dversa
ry
t
hat
interce
pts
a
ci
phertext
a
nd
trie
s
t
o
dec
rypt
it
us
in
g
ra
ndom
keys.
T
he
sta
nd
a
r
d
HE
sch
e
m
e
was
desi
gn
e
d
a
nd
pro
vid
es
pro
va
ble
secur
it
y
only
in
th
e
con
te
xt
of
Pas
swor
ds
,
Pi
ns
a
nd
RS
A
Keys
.
Re
search
ef
f
ort
s
over
t
he
ye
ars
to
exte
nd
i
t
fo
r
e
ncodin
g
la
rg
er
do
m
ai
ns
(s
uch
as
hu
m
an
-
generate
d
m
essage)
has
fail
ed
to
produce
c
onvi
ncin
g
dec
oy
m
essage,
do
not
scal
e
well
.
W
orst
is
they
rev
eal
str
uctu
ral
inf
or
m
at
ion
of
the
un
der
ly
in
g
plaint
ext,
thus,
the
vulne
rab
il
it
y
to
chose
n
ci
ph
e
rtext
at
ta
ck
(CCA)
w
hi
ch
m
akes
the
pro
bab
il
it
y
of
m
essage
rec
ov
ery
hi
gh.
Our
pro
po
se
d
a
ppr
oach
is
the
first
su
cce
ssfu
l
im
ple
m
e
ntati
on
of
ada
pting
the
hone
y
encr
ypti
on
schem
e
fo
r
sup
portin
g
enc
od
i
ng
of
hu
m
an
-
gen
e
rat
ed
m
essage.
It
produces
a
pl
ausible
m
essa
ge
that
is
good
en
ough
to
de
cei
ve
the
ad
ve
rsar
y.
More
im
po
rta
ntly
,
the
crit
ic
al
m
essage
(
ke
ywo
rd
s
)
that
the
sen
der
m
igh
t
wa
nt
to
protect
is
com
plete
ly
con
ceal
e
d
a
nd
cannot
ap
pea
r
durin
g
decr
ypt
ion
with
inc
orr
ect
keys,
ther
e
by,
co
unte
rin
g
the
CC
A.
How
ever,
we
m
us
t
state
that
ou
r
ap
pr
oach
im
pacts
t
i
m
e
wh
en
the
m
essage
gets
la
rg
er.
F
or
th
is
lim
it
a
ti
on
,
we
are
think
i
ng of
othe
r way
s to
c
urt
ai
l t
his in our
ne
xt r
esea
rc
h
e
f
forts.
6.
CONCL
US
I
O
N
In
the
c
onve
nt
ion
al
enc
rypti
on
sc
hem
e
,
m
essages
a
re
en
cod
e
d
in
AS
C
II
(
or
UT
F
-
8
,
et
c.
)
w
hich
m
akes
them
pr
on
e
t
o
a
brute
-
force
at
ta
ck
,
a
s
an
a
dv
e
rsa
ry
who
trie
s
t
o
de
crypt
a
ci
pher
te
xt
can
easi
ly
filt
e
r
ou
t
c
orrect
m
essages
f
ro
m
incorrect
on
es
ba
sed
on
the
non
-
un
if
orm
distribu
ti
on
of
th
e
cha
racters
.
Honey
encr
y
ption
(HE)
offe
rs
a
co
unte
rm
easur
e
to
the
co
nv
e
ntio
nal
enc
ryptio
n
schem
e
by
of
f
erin
g
a
ba
rr
ie
r
agains
t
brute
-
force
at
ta
cks.
H
E
is
a
deco
y
-
base
d
e
ncr
y
ption
sc
he
m
e
wh
ic
h
yi
eld
s
plau
sible
lo
ok
i
ng
plainte
xt
s
under
decr
y
ption
with
inc
orrect
ke
ys
su
ch
t
hat
de
crypti
on
at
te
m
pts
al
on
e
ar
e
insuffici
ent
t
o
rec
over
t
he
correct
plainte
xt.
As
a
res
ult,
an
a
dv
e
rsa
ry
is
m
ade
to
belie
ve
he
has
t
he
plainte
xt
wh
e
n
wh
at
he
ha
s
is
a
deco
y
-
m
essage.
In
this
pa
pe
r,
we
m
od
ifie
d
the
HE
sc
hem
e
to
su
pp
or
t
the
encodin
g
of
hum
an
-
ge
ner
at
e
d
me
ssage
by
le
ver
a
ging
nat
ur
al
la
ngua
ge
processi
ng
te
c
hn
i
qu
e
s.
O
ur
a
ppr
oach
m
od
el
s
the
connecti
on
betwee
n
wor
ds
with
the
sp
eci
fic
go
al
of
ca
pturin
g
the
patte
r
ns
of
la
ngua
ge
to
form
synta
ct
icall
y
and
sem
a
ntica
ll
y
m
eani
ngf
ul
m
essage
as
a
fra
m
ewo
r
k
to
bu
il
d
real
ist
ic
bu
t
fak
e
m
essages
as
deco
ys
w
hich
is
serv
e
d
to
an
ad
ver
sa
ry
try
ing
to
decr
y
pt
the
ci
ph
e
rtext
us
i
ng
an
in
valid
-
ke
y.
Fo
r
in
sta
nc
e,
an
at
ta
cker
that
intercepts
an
encr
y
pted
e
m
ai
l
m
essage
and
tr
ie
s
to
decr
ypt
the
m
essage
gets
a
plausib
le
em
ai
l
m
essage
,
a
nd
he
is
foole
d
into
belie
ving
he
has
the
or
i
gina
l
e
m
a
il
m
ess
age
.
The
m
essage
struct
ur
e
and
m
essage
le
ng
t
h
co
ntaine
d
in
the
unde
rly
ing
m
essage
is
kep
t
entirel
y
secret
and
fail
ed
de
crypti
on
pro
duces
ra
dical
ly
diff
e
re
nt
m
essage
from
the
or
iginal
m
essage
as
d
e
coys
m
essage,
and
su
c
h
feat
ur
es
can
be
use
fu
l
for
s
ucc
essfu
ll
y
co
nce
al
ing
a
nd
prot
ect
ing
e
m
ai
ls and
hum
an
-
w
ritt
en
docum
e
n
ts
that ar
e top secret
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
3
,
June
201
9
:
18
7
1
-
1878
1878
REFERE
NCE
S
[1]
Gam
ido
HV
,
Sis
on
AM
,
Medina
RP
.
,
"
Modifi
e
d
AE
S
for
Te
xt
and
Im
age
En
c
r
y
pt
ion,
"
Indon
e
sian
Journal
of
El
e
ct
rica
l
Eng
in
ee
ring a
nd
Computer
Sc
ie
nc
e
(
IJ
EE
CS)
,
vol
.
11(
3),
pp
.
27
,
Ma
y
2018
.
[2]
Chandra
ka
la
N
,
Rao
BT
.
,
"
Migrat
ion
of
Virt
ual
Ma
chi
ne
to
improve
the
Secur
ity
of
Clo
ud
Com
puti
ng,
"
Inte
rnational
Jo
urnal
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJE
CE)
.
vol
.
8(1)
,
pp.
210
-
9
,
Feb
2
018
.
[3]
Sulta
npur
e
KA
,
Gupta
A,
Redd
y
LS.
,
“
An
Eff
ic
i
ent
Cloud
Sched
uli
ng
Algorit
hm
for
the
Conserva
ti
on
of
Energ
y
through
Broadcasti
ng,
”
Int
ernati
onal
Journal
of
Elec
tri
cal
a
nd
Computer
Engi
ne
ering
(
IJE
CE)
,
vol.
8(1
)
,
pp.
179
-
88
,
Feb
2018
.
[4]
Singh
JP
,
Kum
ar
S.
"
Authen
ti
c
ation
and
encr
y
ption
in
cl
oud
computing
,"
In
Smar
t
Techno
logi
es
a
nd
Manage
men
t
for
Computing,
Comm
unic
ati
on,
Controls,
Ene
rg
y
and
Ma
te
rials
(
ICSTM
)
,
2015
Inte
rnational
Co
nfe
renc
e
on
201
5
May
6
,
IEEE
,
p
p.
216
-
219
.
[5]
Jabe
r
AN
,
Zo
l
kipl
i
MF
,
"
Us
e
of
cr
y
p
togra
p
h
y
in
c
loud
co
m
puti
ng.
In
Co
ntrol
S
y
s
te
m
,
"
Computing
and
Engi
ne
ering
(
ICCSCE)
,
2013
IE
EE
In
te
rnationa
l
Confe
ren
ce on 2013 Nov
29
.
I
E
EE
,
pp.
179
-
184.
[6]
Chen
D,
Zha
o
H.
,
"
Data
sec
ur
i
t
y
and
pr
iva
c
y
prote
c
ti
on
is
sue
s
in
cl
oud
computing
,"
In
Computer
Sci
en
ce
an
d
El
e
ct
ronics
Enginee
ring (
ICCSE
E)
,
2012
Int
ernati
onal
Con
fe
ren
c
e
on
2012
Mar 2
3.
IE
EE,
v
o
l. 1,
pp.
647
-
651
.
[7]
Om
ola
ra
OE,
Ol
udar
e
AI,
Abdul
ahi
SE.
,
"
Deve
l
oping
a
m
odifi
e
d
H
y
brid
Ca
esa
r
ci
pher
and
Vig
ene
re
ci
pher
for
sec
ure
D
at
a
Co
m
m
unic
at
ion
,"
Computer
Engi
n
ee
ring a
nd
Int
el
l
ige
nt
Syste
ms
,
v
ol.
5
,
pp
.
34
-
46
,
2014.
[8]
Stal
li
ngs
W
,
Brown
L,
Baue
r
MD
,
Bhat
ta
ch
ar
je
e
AK
,
"
Com
pute
r
sec
uri
t
y
:
pr
inc
iples
and
pra
ct
i
ce
,"
P
earson
Educ
ati
on
2012.
[9]
Lore
k
P,
Za
gó
rski
F,
Kulis
M.
,
"
Strong
st
at
ion
ar
y
ti
m
es
and
it
s
use
in
cr
y
ptogr
aph
y
,"
arXiv
preprin
t
arXiv
:1709.
0263
1
,
Sep
2017.
[10]
Jo
HJ
,
Yoon
J
W
,
"
A
new
count
e
rm
ea
sure
aga
inst
brute
-
forc
e
at
t
a
cks
tha
t
use
hig
h
-
per
form
anc
e
c
om
pute
rs
for
big
dat
a
an
aly
s
is
,"
I
nte
rnational
Jou
rnal
of
Di
stribu
t
ed
Sensor N
et
wo
rks
.
vol
.
11(6)
,
p
p.
406915
,
Jun 2
015.
[11]
Beuna
rde
au
M,
Ferra
di
H,
G
éraud
R,
Nac
cache
D.
,
"
Hone
y
En
cr
y
pt
ion
for
La
ngu
age
,"
In
Inte
rnat
ional
Conf
ere
nc
e
on
Cryptol
og
y
in
Malay
sia
.
Sprin
ger,
Cham,
pp.
1
27
-
144
,
De
c
201
6
.
[12]
Kim
JI,
Yoon
JW
,
"
Hone
y
cha
t
ti
ng:
A
novel
insta
nt
m
e
ss
agi
ng
sy
st
em
robust
to
ea
vesdropping
over
comm
unic
at
ion
,"
In
Ac
ousti
cs,
Spee
ch
and
Sign
al
Proce
ss
ing
(
ICASSP
)
,
2016
I
EE
E
In
te
rnation
al
Confe
ren
ce
o
n
2016
Mar 20
.
IE
EE
,
pp.
2184
-
21
88.
[13]
Om
ola
ra
AE,
Ja
nta
n
A,
Abiodu
n
OI,
Pos
ton
H
E
,
"
A
Novel
Approac
h
for
th
e
Adapta
t
ion
of
Hone
y
Enc
r
y
p
ti
on
to
Support
Natur
al
La
ngu
age
Mes
sage
,"
In
Proceedi
ngs
of
the
I
nte
rnational
Mu
lt
iConf
ere
nc
e
of
Engi
ne
ers
and
Computer
Scient
ists 2018
,
v
o
l. 1
,
2018
.
[14]
Co
usins
DB,
Ro
hloff
K,
Sum
or
ok
D.,
"
Designing
an
FP
G
A
-
ac
ce
l
era
t
ed
hom
o
m
orphic
enc
r
y
pt
ion
co
-
proc
essor
,
"
IEE
E
Tr
ansacti
o
ns on
Eme
rging
Topics
in
Comp
uti
ng
.
vol. 5(2)
,
pp.
193
-
206
,
Ap
r
2017
.
[15]
Kestur
S,
Davis
JD
,
W
il
liam
s
O
,
"
Bla
s
compa
rison
on
FPGA
,
CP
U
and
GPU
.
VLSI
(ISV
LSI),
"
2010
I
EE
E
comput
er
soci
ety
annual
sympos
i
um on
2010
Jul
5.
IE
EE,
pp
.
288
-
293.
[16]
Sutikno
T,
Idris
NR,
Jidin
A.
,
"
High
-
Speed
Com
puta
ti
on
using
FP
GA
for
Exc
el
le
nt
Perform
ance
of
Dire
ct
Torq
ue
Control
of
Ind
uct
ion
M
ac
hin
e
s,"
Tele
commun
ic
ati
on
Comput
ing
Elec
tronic
s
and
Control
(
TEL
KOMNIKA)
,
vol.
14(1)
pp.
1
-
3,
Mar
2016
.
[17]
Marks
M,
Jantu
ra
J,
Ni
ewia
dom
ska
-
Sz
y
nki
ewicz
E,
Strz
el
c
z
y
k
P
,
Góźdź
K.
,
"
He
te
roge
n
eous
GP
U&CP
U
cl
uster
for
high
-
per
form
anc
e
computing
in
cr
y
p
togra
ph
y
,
"
Computer
Sc
ience
,
vol
.
13
,
pp
.
63
-
79,
2012
[18]
Couture
N,
Ken
t
K
B
,
"
The
eff
e
ct
iv
ene
ss
of
brut
e
forc
e
attac
ks
on
RC4
,"
Comm
unic
ati
on
N
et
wo
rks
and
Serv
ic
es
Re
search, 2004.
Proce
ed
ings.
S
e
cond
Annua
l
Co
nfe
renc
e
on
200
4
May
19
.
IEEE
,
pp.
333
-
336.
[19]
Juels
A,
Riste
npar
t
T.
,
"
Hone
y
enc
r
y
p
ti
on:
Se
cur
ity
be
y
ond
t
he
brute
-
forc
e
bo
und
,"
In
Annual
Inte
rnationa
l
Confe
renc
e
on
the
Theory
and
Appl
ic
at
ions
of
Cryptographic
Techni
ques
2014
May
11
,
Springer,
Be
rlin
,
Heide
lb
erg
,
pp
.
293
-
310.
[20]
Juels
A,
Riste
npar
t
T.
,
"
Hone
y
E
ncr
y
p
ti
on:
En
cr
ypti
on
be
y
ond
th
e
brute
-
forc
e
bar
rie
r
,"
IE
EE
Sec
u
rity
&
P
rivac
y
,
vol.
12(4)
,
pp
.
5
9
-
62,
Jul
2014
.
[21]
Bird
S,
Lope
r
E.
,
"
NLTK:
th
e
na
tura
l
la
nguag
e
to
olki
t
,"
In
Procee
dings
of
the
AC
L
2004
on
Inte
ractive
poster
an
d
demonstration
sess
ions 2
004
Jul 21
,
Association
for Computat
ion
al
Linguisti
cs
,
p
p
.
31
.
[22]
Chowdhur
y
G
G.
,
“
Natur
a
l
l
angu
age
pro
ce
ss
ing
,”
Annual
r
ev
i
ew
of
inf
orm
ati
on
scie
n
ce
and
t
ec
hn
ology
,
vol.
37(1)
,
pp.
51
-
89
,
Jan
2
003
.
Evaluation Warning : The document was created with Spire.PDF for Python.