0% found this document useful (0 votes)
31 views

Color Image Compression-Encryption Algorithm Based

Uploaded by

Apara
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
31 views

Color Image Compression-Encryption Algorithm Based

Uploaded by

Apara
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2019.Doi Number

Color Image Compression-Encryption


Algorithm Based on Fractional-Order Memristor
Chaotic Circuit
1 1 2 1 1
FEIFEI YANG , JUN MOU , KEHUI SUN , YINGHONG CAO , and JIYU JIN
1
School of Information Science and Engineering, Dalian polytechnic University, Dalian, 116034, China
2
School of Physics and Electronics, Central South University, Changsha, 410083, China
Corresponding author: Jun Mou (e-mail: moujun@csu.edu.cn).
This investigate is supported by the Basic Scientific Research Projects of Colleges and Universities of Liaoning Province (Grant Nos. 2017J045);Provincial
Natural Science Foundation of Liaoning (Grant Nos. 20170540060); Scientific Research Projects in General of Liaoning Province (Grant Nos.L2015043)
and Doctoral Research Startup Fund Guidance Program of Liaoning Province (Grant Nos. 201601280)

ABSTRACT In this paper, a fractional-order memristive chaotic circuit system is defined according to
memristor circuit. The dynamic characteristics are analyzed through phase diagram, bifurcation diagram
and Lyapunov exponent spectrum, and the randomness of chaotic pseudo-random sequence is tested by
NIST SP800-22. Based on this fractional-order memristive chaotic circuit, we propose a novel color image
compression-encryption algorithm. In this algorithm, Compression Sensing (CS) algorithm is used for
compression image, and then using Zigzag confusion, add modulus and BitCircShift diffuse encrypt image.
Theoretical analysis and simulation results indicate that the proposed compression and encryption scheme
has good compression performance, reconstruction effect, and higher safety performance. Moreover, it also
shows that the new algorithm facilitates to encryption, storage and transmission of image information in
practical applications.

INDEX TERMS Color image encryption, Compression Sensing (CS), Zigzag confusion, Add Modulus and
BitCircShift diffuse, Fractional-order memristive chaotic circuit

I. INTRODUCTION Wu et.al designed a lossless color image encryption


As the development of science and technology, the algorithm, the scheme by 6D hyperchaotic and plane-image
information safety becomes important. Compared with other to improve safety performance [9]. The image encryption
information carriers, color image has more information. scheme according to block confused and dynamic index
Therefore, it is necessary for us to study color image diffused is introduced by L et al. [10]. These different image
encryption. encryption algorithms [4-8, 10] just encrypted image, and not
Due to the inherent performance of chaotic system with compress image, which bring high storage and transmission
randomness, non-periodicity and sensitivity to parameters costs of the information. Because compression sensing (CS)
and initial values [1-3], it is widely applied to image can effectively compress and encrypt image, therefore, we
encryption algorithm. At the moment, various color image propose a novel color image compression-encryption
encryption algorithms using chaotic system were proposed algorithm through CS to overcome these shortcoming.
[4-21]. Huang et al. applied the unpredictable characteristics Compared to other chaotic system, the fractional-order
of chaotic system to encrypted color image [4]. Choquet chaotic system possesses more abundant dynamics
fuzzy integral and piecewise linear of chaotic map were characteristics, because of its nonlocal features and high
applied to encrypted color image [5]. Wei et al. [6] nonlinearity [22-25]. What’s more, the fractional-order
introduced a color image encryption scheme through DNA chaotic system is used to cryptosystem, which increase the
sequence operation and chaotic sequence. Kadir et al. key space and improve security. The fractional-order
represented a color image encryption algorithm by skew tent memristive system is solved through Adomian
map and hyperchaotic system [7]. A new color image decomposition method (ADM) algorithm [26-28], which
encryption algorithm is presented through Rhouma et al. [8]. reduces computation time and enhances space complexity. In

VOLUME XX, 2019 1


2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2017)

addition, memristor chaotic system is easy to implement in coefficients not equal to non-zero and (N-K) coefficients
practical application. Memristor, a nonlinear two-terminal equal to zero, y is deemed to be sparse. The sparse signal y is
electronic component was predicted by Chua [28]. It is used measured through the measurement matrix Φ∈RM×N,and
for construction chaotic oscillators through replacing linear the corresponding measured value Y is obtained by
or nonlinear resistance components [29-33]. Until 2008, the
Y  y  S  S , (2)
memristor was fabricated in I Iewlett-Packard Lab [34],
which shown that the memristor was physically realized. where Θ represents a Μ×N matrix.
After that, many new memristive chaotic systems were The signal reconstruction is essentially a linear equation
proposed and practical circuit could be realized [35-37]. solving process. Because the number of unknowns is more
Therefore, in this algorithm, we chose the fractional-order than equations in Eq. (2) and the coefficients S are sparse, the
memristive chaotic circuit for image encryption. Eq. (2) has a group of multiple solutions. The minimum
In 2006, Candes and Donoho proposed CS theory. They norm l0 to reconstruct signal can be solved, if the
proved that sample signal much smaller than the data measurement matrix Φ and the basis matrix Ψ are meet to
sampling rate specified by Nyquist’s theorem when the (Restricted Isometry Property) RIP [43]. For all y∈ ∑k, the
signal is sparse or compressible [38], and the high probability existing δk∈(0,1) is used to:
signal can be accurately reconstructed. To improve security
(1   k ) y 2  Ay 2  (1   k ) y 2 ,
2 2 2
and compression features of the image compression and (3)
encryption algorithm has been proposed, an improved
algorithm by CS and 4D hyperchaotic system is proposed by where y is k-order sparse signal. δk is RIP constant, and
Tong et al. [39]. Zhu et al. [40] designed a novel matrix A∈RM×N is meet k-order RIP.
compression-encryption algorithm by 2D discrete The signal can be accurately reconstructed by
hyperchaotic system, CS and Chinese remainder theorem. A 

color image encryption algorithm through CS and fractional S  arg min S 0 , s.t.Y  S , (4)
Fourier transform is proposed [41]. Chai et al. [42] proposed
where ‖•‖0 is vector norm l0, and it means that the amount
a visually secure image encryption scheme based on
of non-zero elements in the vector y.
compressive sensing. Among them, they just performed one
We use discrete cosine transform (DCT) to extend image
time compression, which would not enough to cut short the
pixel matrix. The measurement matrix is obtained by chaotic
storage and transmission costs of encrypted image. To
pseudo-random sequence and Hadamar matrix. The
adequately reduce the costs, our algorithm performs twice
orthogonal matching pursuit (OMP) algorithm is applied to
compression. In addition, the chaotic pseudo-random
reconstruct image.
sequences were not tested randomness even though they used
in these algorithms. In this paper, the randomness of chaotic
B. ZIGZAG CONFUSION
pseudo-random sequences is tested by NIST SP800-22.
Zigzag confusion refers to the transformation process of
In this paper, we focus on investigating a color image
starting from the upper left element of the matrix, scanning
compression-encryption algorithm through CS and
the element in the matrix in the order of Z, and reorganizing
fractional-order chaotic circuit. The rest of the paper is
them into the same size matrix in line. It is also called the
organized as follows. In Section 2, the basic definition and
standard transformation. Generally, 4×4 or 8×8 sub-blocks
preliminary are introduced. In Section 3, dynamical
are used as templates, or the whole square matrix can be
behaviours of fractional-order memristive chaotic circuit are
scanned directly. For example, the zigzag confusion of a 4×
analyzed, and its pseudo-random sequences are designed and
4 matrix is shown in Fig. 1. Where the Fig. 1(a) is original
tested. The proposed algorithm is described in section 4. In
matrix, the start pixel’s position is (1, 1) of zigzag confusion
section 5, the simulation results of the proposed algorithm are
process as shown in Fig.1 (b), and the result of matrix zigzag
given. The security performances are analyzed in section 6.
confusion is shown in Fig.1(c). From the Fig. 1, we can see
Finally some conclusions are obtained.
that the matrix can be effectively scanned by zigzag
II. BASIC DEFINITION AND PRELIMINARY
confusion.

A. COMPRESSION SENSING (CS)


Assumption one dimensional signal y =[y(1), y(2),…, y(N)]T,
the linear combination of its orthogonal basis is defined by
N
y   S   Si  i , (1)
i 1
(a) (b) (c)
where Ψ, S are basis matrix, column vector, N×1 column FIGURE.1 Zigzag confusion (a) original matrix, (b) zigzag confusion, (c)
vector and the sparse coefficient of y respectively. If S has K the matrix after zigzag confusion

2 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2017)

C. ADOMIAN DECOMPOSITION METHOD (ADM)


 0 (t  t0 ) k
 k 0 k k !
m 1
Suppose the fractional-order chaotic system *Dqto(t)=f(x(t))+
 x  J q
t0 g  b
g(t), here x(t)=[x1(t),x2(t),…,xn(t)]T are given function variab 
les, in autonomous systems, g(t)=[ g1(t),g2(t),…,gn(t)]T are c  x1  J tq0 Lx 0  J tq0 A0 ( x 0 )
onstants, f(x(t)) including linear and nonlinear partial functi  2
 x  J t 0 Lx  J t 0 A ( x , x )
q 1 q 1 0 1
ons. The system is divided into three parts by the following . (12)
formulation [22], ...

*Dtoq x(t )  Lx(t )  Nx(t )  g (t )  xi  J tq0 Lxi 1  J tq0 Ai 1 ( x 0 , x1 ,..., x i 1 )
 (k )  
 x (t0 )  bk , k  0,1,..., m  1 , (5) ...
m  N , m  1  q  m
 III. FRACTIONAL-ORDER MEMRISTOR CHAOTIC
q CIRCUIT PSEUDO-RANDOM SEQUENCES
where *D means that the derivative operator of order q, L
to

represent the linear part of the system function and N is the


A. FRACTIONAL-ORDER CHAOTIC CIRCUIT
non-linear part of system function, and bk is initial value.
Here, the memristor chaotic circuit as shown in Fig. 2(a), the
By applying the J qto both sides of Eq. (3) and obtained
following equation [23]: equivalent circuit of voltage-controlled memristor is shown
in Fig. 2(b). On this basis, a fractional-order memristive
m 1
(t  t0 )k chaotic circuit is defined.
x  J tq0 Lx  J tq0 Nx  J tq0 g   bk , (6)
i2
k! R1 L + i
k 0 +
U -
where Jqto represent the integral operator of order q, and t0≤t + +
v2 v1 v
R2
≤t1,q≥0,r≥0,γ>-1 and C is real constant. The basic C2 - - C1
R W
characteristics of the integral operator Jqto are presented by -
[22]:
(  1) (a)
J tq0 (t  t0 )  (t  t0 )  q , (7)
(  1  q) + i - v0 + Rc
v
Ua Mb Rb +
C - Ra C0 - Uc
+ -
J C
q
(t  t0 )q , (8) - +
(q  1)
t0
Ub Ma Rd

(b)
J tq0 J tr0 x(t )  J tq0r x(t ) . (9)
FIGURE.2 Memristor circuit based on Chua’s circuit, (a) simple circuit
schematic, (b) Voltage-controlled memristor circuit
According to principle of Adomian decomposition
Based on Fig. 2(b), the relationship of input voltage v and
algorithm, the non-linear parts of Eq. (5) are decomposed
current i, and voltage v0 and capacitor C0, are described as
by
follows
 i 1 di
 Aj  [ i N ( j ( ))] 0
i dv0 1
 C0  v, (13)
 i ! d , (10) dt Ra
 ( ) 
 j
i
 k  0 ( ) x j
i k k

1
i  W (v0 )v   (1  gv02 )v , (14)
where, i∈(0,∞),j∈(1, n). Then the non-linear parts are Rb
express as

where g represents overall gain of multipliers Ma and Mb.
Nx   A ( x , x ,..., x ) .
i 0 1 i
(11)
For the variables v0, v1, v2 and i in Fig. 2(a), the circuit
i 0
equations are given by

So the solution of Eq. (5) x=∑∞i=0 is described by

2 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2017)

 dv0 1  A0bgx ( x )  bgx20 ( x10 )2


C0 dt   R v1
2

 2 1

 a  A1 bgx ( x ) 2  bg ( x12 ( x10 ) 2 +2 x20 x11 x10 )


 dv1 1  2 2 1

C1 dt  R (1  gv0 )v1  i2  Abgx ( x )  bg ( x22 ( x10 ) 2  2 x12 x11 x10  2 x20 x12 x10  x20 ( x11 ) 2 )
2
2

 3
2 1

 a , (15)
 Abgx ( x )  bg ( x23 ( x10 ) 2  2( x22 x10 x11  x20 x13 x10  x20 x11 x12 ))
 dv2 v2 
2

C2 dt  R  i2
2 1

 bg (2 x12 x12 x10  x12 ( x11 ) 2 )


  4
 L di2  v  v  Abgx ( x ) 2  bg ( x24 ( x10 ) 2  2( x23 x10 x11  x22 x12 x10  x12 x11 x12 ))
 dt 
2 1
2 1
 2bg (2( x12 x13 x10  x20 x14 x10  x20 x13 x10 )  x22 ( x11 ) 2
where v=v1.   x20 ( x12 ) 2 )
Let x=v0, y=v1, z=v2, w= Ri2, τ=t/RC2, Rb=R, a=RC2/RaC0, 
b=C2/C1, c=R2C2/L, so the Eq. (15) is defined as [44] . (19)
The initial conditions are
 x  ay
  1 0 (t  t0 )
q
 y  b(1  gx ) y  bw
2
 1
x   ax
 , (16) 2
(q  1)
z  z  w 
 w  c( z  y )  1 0 (t  t0 )
q

 x2  (b( x2  gx2 ( x1 ) )  bx4 )


0 0 0 2

 (q  1)
where a, b, c and g represent system parameters. Set  , (20)
 x1  ( x 0  x 0 ) (t  t0 )
q
a=12.375, b=7.0213, c=2.475, g=0.2 and x∈(0, 1.5, 0, 0), we
get the Lyapunov exponents (0.0805, 0, -0.0303, -34.5343).  3 3 4
(q  1)
Obviously, in these Lyapunov exponent values, there is a 
positive exponent, which shows that the system is chaotic.  1 (t  t0 ) q
 x4  (b( x3  x2 )) (q  1)
0 0
According to the fractional-order definition, and
corresponding the fractional-order system is given by

where xj0 is the initial values of system (17), h=t-t0, and then
 D x  ax2
c q
t0 1 the solution of system (17) can be represented by
c q
 Dt0 x2  b(1  gx1 ) x2  bx4
2
(t  t0 ) q (t  t0 ) 2 q
c q , (17) x j (t )  c 0j  c1j  c 2j
 Dt0 x3  x3  x4 (q  1) (2q  1)
, (21)
c q (t  t0 )3q (t  t0 ) 4 q
 Dt0 x4  c( x3  x2 ) c3j  c 4j
(3q  1) (4q  1)
where q represent order of the equation. When q=1, system
(17) becomes system (16). where

B. THE SOLUTION OF FRACTIONAL-ORDER


c10  x10
 0
MEMRISTOR CHAOTIC CIRCUIT
c2  x20
The fractional-order system (17) is derived by ADM  0 , (22)
algorithm, and then linear and nonlinear terms are obtained c3  x30
as follows c 0  x40
 4
 Lx1   -ax2   N x1  0 
 L   bx  bx   N    c11  ac20
 x2  =  2 4   x2   bgx1 x2 
2
, = . (18)  1
 Lx 3   x3 - x4   N x 3  0  c2  bc2  bc4  bgc2 (c1 )
0 0 0 0 2

         1 , (23)
 Lx 4  c( x3 - x2 )   N x 4  0  c3  c3  c4
0 0

c1  c(c 0  c 0 )
Based on Eq. (10), the before five Adomian polynomials  4 3 2
for the nonlinear part –bgx12x2 are

2 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2017)

c12  ac12 FIGURE.3 Dynamic characteristics of the fractional-order memristor

 2 chaotic system, (a) x-y plane, (b) bifurcation diagram for q∈(0.7, 1), (c)
c2  bc12  bc14  bg(c12 (c10 )2 +2c20c11c10) Lyapunov exponents spectrum for q∈(0.7, 1)
 2 , (24)
c3  c31  c14 D. DESIGN OF FRACTIONAL-ORDER CHAOTIC
c 2  c (c  c )
1 1 PSEUDO-RANDOM SEQUENCE
 4 3 2 The quantized sequences of chaotic sequences are called
chaotic pseudo-random sequences. The randomness of
c13  ac22 chaotic system is mostly reflected in the performance of
 3 quantized random sequences. Quantization of chaotic real
c2  bc2  bc4  bg(c2 (c1 ) +2c2 c1 c1
2 2 2 0 2 0 2 0
value sequence is an important part of generating pseudo-
 (2q  1)
 random sequence design. This process directly affects the
+(2c2 c1 c1 +c2 (c1 ) )) 2
1 1 0 0 11 2
, (25)
  (q  1) randomness, complexity and other characteristics of
sequences, and ultimately affects the security of its
c 3  c 2  c 2 application system. Set a=12.375, b=7.0213, c=2.475, g=0.2,
 3 3 4
initial system value (x1, x2, x3, x4 )∈(0, 1.5, 0, 0), then the
c4  c(c3  c22 )
3 2
system are decimals as the chaotic sequences. The specific
steps as follows.
c14  ac23 Step.1 After the system parameters and initial values are
 4 determined, the system was iterated for 5000 times to
c2  bc23  bc43  bg
((c22 (c10 ) 2 +2c20c13c10 +
eliminate the transient effect, and then fractional-order
 (3q  1) memristive chaotic circuit system is continued to be iterated.
 2(c22 c11c10 +c12 c12 c10  c20 c11c12 )
 (q  1)(2q  1) We get the four decimals sequences, and then new four
 . (26) decimals sequences are generated by
 (3q  1)
+ c (c ) 3
3 11 2
) max( x)  min( x)
 2
 (q  1)
1 
 4  x1( j )  2  x( j )  max( x)  min( x)
c3  c33  c43 
c 4  max( y )  min( y )
 4  c(c33  c23 )  y1( j )  2  y ( j )  max( y )  min( y )

 , (27)
C. DYNAMICAL ANALYSIS OF FRACTIONAL-ORDER
CHAOTIC CIRCUIT
 z1( j )  2  z ( j )  max( z )  min( z )
Here, set the system parameter a=12.375, b=7.0213,
 max( z )  min( z )

c=2.475, g=0.2, q=0.98, h=0.001 and initial values (x1, x2,
 w1( j )  2  w( j )  max( w)  min( w)
x3, x4 )∈(0, 1.5, 0, 0), and we get Lyapunov exponents  max( w)  min( w)
(0.1209, 0, -0.0580, -49.9689), Lyapunov dimension
DL=3.0013 and phase diagram as shown in Fig.3 where x(j), y(j), z(j) and w(j) represents four chaotic
(a).Therefore, the system is chaotic. Then let the other sequences, max and min are Maximum and Minimum of
parameters and initial values keep unchanged, order chaotic sequence. In addition, x1(j), y1(j), z1(j) and w1(j)
q∈(0.7, 1), bifurcation diagram and Lyapunov exponents denote new decimals sequences.
spectrum are shown Fig.3 (b) and (c). From the Fig. 3(b) Step.2 Decimals sequences are converted to integer
and (c), we can see that the fractional-order memristor sequence by taking the module, integer up and integer down.
chaotic circuit system has better randomness and
parameters sensitivity. Hence, it can improve security
performance to apply in image encryption algorithm.
8 2
Lyapuov esponents

1 6
1
4
0
0 2
x

-1
z

0
~ ~
-1 -2 ~ ~
-50
-4
-250
-2 -6 0.7 0.75 0.8 0.85 0.9 0.95 1
-8 -4 0 4 8 0.7 0.75 0.8 0.85 0.9 0.95 1
x q q
(a) (b) (c)

2 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2017)

 x 2( j )  mod(round (1000  x1( j ) 1016 memristive chaotic circuit passed 15 test indexes of NIST SP
 800-22 random number inspection standard. Moreover, the
  floor ( x1( j ) 1016 )) image encryption algorithm using fractional-order memristor
 chaotic circuit has more high security features
 y 2( j )  mod(round (1000  y1( j ) 1016
 IV. IMAGE COMPRESSION-ENCRYPTION AND
  floor ( y1( j ) 1016 ))
DECRYPTION ALGORITHM

 , (28)
A. COMPRESSION-ENCRYPTION ALGORITHM
 z 2( j )  mod(round (1000  z1( j ) 10
16
The proposed color image compression and encryption

  floor ( z1( j ) 1016 )) algorithm process is shown in Fig.4, and the main process is
consist of decomposition of the image plane, discrete cosine

 w2( j )  mod(round (1000  w1( j ) 10
16 transform (DCT), CS, Zigzag scrambling and add modulus
and BitCircShift diffusion. The details of compression and

  floor ( w1( j ) 1016 )) encryption scheme are as follows.

where x1(j), y1(j), z1(j), w1(j) represents new four decimals Original image
sequences, and x2(j), y2(j), z2(j), w2(j) denote new integer
sequences. B G R

E. NIST SP800-22 TEST DCT DCT DCT


In this paper, the quantized chaotic pseudo-random
sequences are used to image compression and encryption CS CS CS
algorithm, and random sequence performance is analyzed by
NIST SP800-22. For this test, all testing is done Zigzag Zigzag Zigzag Chaotic system
automatically through the test package STS. There are 15 test
indicators and 2 kinds of judgment basis (P-value, pass rate) Mod and Mod and Mod and
in NIST SP800-22[45]. For the pass rate, given significant BitCircShift BitCircShift BitCircShift
level of α, test sequence β group, the confidence interval of
pass-through rate is defined as: Encrypted image

(1    3  (1   ) / ,
1    3  (1   ) /  ) . (29) FIGURE.4 Flowchart of the process encryption algorithm

If the pass rate falls within this confidence interval, which Step.1 The plain color image I with size of H×W is
indicates that the sequence passed the test, otherwise, the test inputted.
is failed. For P-value, if P-value > 0.0001, which shows that Step.2 The plain color image I are decomposed into three
the sequence is random, otherwise, the sequence is not plains R, G and B with size of H×W.
random. Step.3 To sparse the pixel matrix of image, three sparse
Setting the significant level of α=0.01, test sequence matrices of planes R, G and B are obtained by DCT
β=100, each group leader is 1 000 000 bit, the confidence operation. The definition of DCT is as follows,
interval is [0.96, 1], we get the pseudo-random sequences of H 1 W 1
 (2h  1) p  (2w  1)q
test results for fractional-order memristor chaotic circuit as Bpq   p q  Ahw cos cos
shown in Table.1. From the Table.1 we can see that the h  0 w 0 2H 2w
pseudo-random sequence generated by fractional-order , (30)
TABLE.1
TEST RESULTS OF NIST SP 800-22
The test serial number The test name P-value Pass rate Test number Test results
1 Frequency 0.867692 0.99 1 success
2 Block Frequency 0.77918 1 1 success
3 Cumulative Sums (*) 0.739918 0.99 2 success
4 Runs 0.779188 0.98 1 success
5 Longest Run 0.055361 1 1 success
6 Rank 0.474986 0.99 1 success
7 FFT 0.062821 1 1 success
8 Non-overlapping Template(*) 0.071177 0.99 148 success
9 Overlapping Template 0.013569 0.99 1 success

2 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (February 2017)

10 Universal 0.108791 0.99 1 success


11 Approximate Entropy 0.759756 1 1 success
12 Random Excursions (*) 0.249284 0.98 8 success
13 Random Excursions Variant(*) 0.025193 0.98 18 success
14 Serial(*) 0.137282 0.98 2 success
15 Linear Complexity 0.224821 0.97 1 success
Where, *tests contain multiple tests, listed as the worst case.
 1 away the former n values to enhance initial value sensitively,
 H , p0 we can get four chaotic sequences, and then four pseudo-

p   , (31) random sequences are generated by Eq. (27) and Eq. (28).
Step.9 In diffusion operations, the pixel values are diffused
 2
, 1  p  H 1 by new pseudo-random sequence. The new two pseudo-
 H
random sequences S1 and S2are obtained by
 1  p(4 j - 3)  x( j )
 , q0
  p(4 j - 2)  y ( j )
q  
W 
, (32)  , (34)
 2  p (4 j -1)  z ( j )
, 1  q  W 1
 W  p(4 j )  w( j )
where A is pixel matrix of original image, and B represents where x(j), y(j), z(j) and w(j) represent four pseudo-random
the new matrix, H and W means that the length and height of sequences, and p represent new pseudo-random sequence.
the image. Step.10 The diffusion operations for three scrambled
Step.4 To get the measurement matrix, for the fractional- image pixel matrices are performed through the
order memristor chaotic circuit system, setting the parameter Modularization and the BitCircShift algorithm. The diffusion
and initial values, let the system (20) iterate m+M times and results are obtained by
thrown away the former m values to enhance initial value
sensitivity. We obtain four chaotic sequences, and then the Ci  (Ci 1  Si  Pi )  LSB3 (Ci 1 )
four pseudo-random sequences are obtained by quantization  , (35)
operation of chaotic sequences, the quantitative principle is Ci  (Ci 1  Si  Pi ) mod 256  LSB3 (Ci 1 )
defined in section III D.
where Si are the pseudo-random sequences S1 and S2. LSB3
Step.5 Three pseudo-random sequences are randomly
represents the lowest three digits of the data, and the
selected from the Step.4, and then three measurements
proposed algorithm using the 8b image, therefore, each pixel
matrices of M×M are obtained by combining the Hadamar
has 8b, the lowest three digits of the any one data is (0, 7).
matrix and M=CR(compression ratio) × H. Here, Hadamar
This is the valid range of circular bits for a pixel point data.
matrix is generated by
Step.11 The encrypted image is obtained through
 1 1 combining three diffused image pixel matrices.
1 1 . (33)
  B. DECRYPTION ALGORITHM
Encrypted image
Step.6 According to mensurement, three sparse plain
matrices after DCT is calculated by compression using three B G R
measurement matrices. We get three compressed image pixel
matrices H × M, on the basis of this, by calculating the IMod and
IBitCircShift
IMod and
IBitCircShift
IMod and
IBitCircShift
second mensurement matrix. Then three compressed image
pixel matrices M×M are obtained. IZigzag IZigzag IZigzag
Step.7 The confusion operations of the three compressed Fractional-order
chaotic system
pixel matrices. The pixel positions of three compressed OMP OMP OMP
image pixel matrices are scrambled by the zigzag confusion
in Section 2.2, and we can get three confused image pixel IDCT IDCT IDCT
matrices.
Step.8 In order to obtain the random of chaotic sequence B G R
of diffusion operations, setting parameters and initial
Decrypted image
conditions of the fractional-order memristor chaotic circuit
system, and let the system (20) iterate n+M times and thrown FIGURE.5 Flowchart of the process decryption algorithm

7 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (March 2019)

It is obvious from the Fig.5 that decryption algorithm process


including the Modularization and the BitCircShift, Zigzag
and DCT inverse algorithm. A detailed process of the
reconstruction and decryption algorithm is presented as
follows.
Step.1 For an encrypted image M×M, decompose it into
three pixel matrices M×M.
Step.2 The Modularization and the BitCircShift inverse (a1) (a2) (a3)
algorithm are used to recover three diffused pixel matrices.
The diffusion pseudo-random sequences are obtained by
Step.8 and 9 of the encryption.
Step.3 Three compressed pixel matrices are regained
through Zigzag reverse algorithm.
Step.4 Three sparse plane matrices N×N are reconstructed
based on OMP algorithm. Where, the measurement matrices
are generated by using Eq. (30) and (31).
(b1) (b2) (b3)
Step.5 The decryption image is obtained by DCT inverse
algorithm and IDCT is calculated as follow FIGURE.6 Algorithm test results, (a1) original Lena image, (a2)

 (2m  1) p  (2n  1)q


M 1 N 1 encrypted Lena image, (a3) decrypted Lena image, (b1) original pepper

Amn   p q Bpq cos cos image, (b2) encrypted pepper image, (b3) decrypted pepper image
m0 n 0 2M 2N
, (36) B. THE COMPRESSION RATIO ANALYSIS
The compression performance of the new algorithm is
 1 analyzed by Mean Structural Similarity (MSSIM) and Peak
 M , p0 Signal to Noise Ratio (PSNR) with different compression

p   , (37) ratios (CR). The CR is defined as [46];
 2 , 1  p  M 1 CM  C N
 M CR  , (39)
IM  IN
 1
 , q0 where IM and IN are length and height of original image. CM
 N and CN are length and height of encrypted image,
q   , (38)
respectively.
 2
, 1  q  N 1
 N
1) MEAN STRUCTURAL SIMILARITY (MSSIM)
The MSSIM is used to estimate the characteristic of the
where B represents the pixel matrix reconstructed by OMP encryption algorithm, and it is described by [47]
algorithm. A denote new IDCT result matrix, M and N means 2 X Y  L1
that length and height of the image. l( X ,Y )  , (40)
 X2  Y2  L1
V. SIMULATION RESULTS
2 X  Y  L2
c( X , Y )  , (41)
A. THE RESULTS OF ENCRYPTION AND DECRYPTION  X2   Y2  L2
ALGORITHMS
Setting the parameters a=12.375, b=7.0213, c=2.475, g=0.2,  XY  L3
h=0.001, q=0.98, x0=0, y0=1.5, z0=0, w0=0, m=500 and n=500, s( X , Y )  , (42)
and then input color “Lena” and “pepper” images with the  X  Y  L3
size of 256×256. The proposed compression and encryption
algorithm are performed in MATLAB 2014a, and the
SSIM(X ,Y )=l (X ,Y )  c(X ,Y )  s(X ,Y ) , (43)
compression ratio is 0.8, the encrypted image and decrypted M
1
image are as shown in Fig. 6. It can be seen from the Fig. 6
that the new algorithm can compression and encryption
MSSIM( X , Y ) 
M
 SSIM( x , y ) ,
k 1
k k (44)

image effectively.
where μ represent the mean of image structural. σ means that
image variance structural. L1, L2 and L3 represent small
constants to avoid denominator of 0. M represent the overall

11 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (March 2019)

number of image blocks. Here, parameters L1=(K1×C)2, which shows that the new algorithm has larger key space and
L2=(K2×C)2, L3=L2/2, K1=0.001, K2=0.003, M=64, C=255. can prevent the brute force attacks. Key space compared
The MSSIM values with different CR are shown in Table.2. results with other algorithms as shown Table.4.
According to the Table.2, the MSSIM values change when TABLE.4
KEY SPACE OF DIFFERENT ALGORITHM
the compression ratio varies, which effectively compress and
Our algorithm Ref.[5] Ref.[6] Ref.[11] Ref.[14] Ref.[19]
encrypt image according to different practical application
2448 2324 2233 2148 2256 2280
needs.
TABLE.2
MSSIM VALUES UNDER THE DIFFERENT CR B. KEY SENSITIVITY ANALYSIS
Lena pepper The sensitively is an important indicator to evaluate the
CR R G B R G B security performance of encryption algorithm. To test the key
0.8 0.7135 0.6551 0.4931 0.7410 0.6437 0.5112 sensitivity, we make a small change in the value of the secret
0.4 0.4478 0.4135 0.4594 0.5191 0.4282 0.5388 key, the decrypted image results as Fig.7. It can be seen that
0.2 0.3290 0.3456 0.3461 0.3563 0.2759 0.3375 if the key value is slightly changed, the decrypted image is
2) PEAK SIGNAL TO NOISE RATIO (PSNR) entirely different from the original Lena image. Moreover,
PSNR is used to evaluate the performance of reconstruction the proposed algorithm is extremely sensitivity to its key.
algorithm to restore image. The formula for calculating
PSNR is defined as [47-48]:

1 r c
MSE   ( Eij  eij )2 ,
rc i 1 j 1
(45)

2552
PSNR  log10 ( ), (46) (a1) (a2) (a3)
MSE
where r and c are length and height of the image, MSE
means that mean error between the original and the restore
image. Eij and eij are pixels of original and restore image in
(i, j) position. The larger of PSNR value shows that the
reconstructed image is closer to the original image. The
PSNR values of different CR are listed in Table.3. The (b1) (b2) (b3)
PSNR values are close to 30 when CR>0.4. Therefore, the
reconstruction result is also good for a small of
compression.
TABLE.3
PSNR VALUES UNDER THE DIFFERENT CR
Lena pepper
CR R G B R G B
0.8 33.370 31.966 30.921 33.787 31.765 30.479 (c1) (c1) (c3)
0.4 30.092 29.581 30.409 30.635 28.184 30.381
FIGURE.7 key sensitively test, (a1) x0+10-15, (a2) y0+10-15, (a3) z0+10-
0.2 28.990 28.627 29.273 29.206 28.434 28.685
15, (b1) w0+10-15, (b2) a+10-15, (b3) b+10-15, (c1) c+10-15, (c2) g+10-15,
(c3) q+10-15
VI. PERFORMANCE ANALYSIS
For the security performance analysis of existing image C. STATISTICAL ANALYSIS
encryption algorithms, Lena image is generally used as the In this section, the statistical performance is analyzed by
target image for performance analysis. In order to facilitate histogram and the correlation coefficient.
comparison, we only analyze the features of Lena image.
1) HISTOGRAM ANALYSIS
A. KEY SPACE
The histogram indicates that the image pixel distribution. For
For a good encryption algorithm, it should be has enough the original image, its pixel distribution fluctuate, on the
large key space and to resist brute force attacks. Our key of contrary, the encrypted image’s pixel distribution is uniform
image compression and encryption scheme is comprised of and flat. The histogram of original and encrypted Lena image
chaotic system parameters a, b, c, g, q, h, initial values x0, y0, with R, G and B are represented in Fig.8. It shows that the
z0, w0, iterations m and n. If the computational accuracy is 10- histogram distribution of original image is fluctuates,
15
, the key space of the proposed algorithm would be 2448, histogram distribution of encrypted image is flat. Moreover,
the proposed algorithm can resist histogram attack.

11 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (March 2019)

1000
900 M
1
  x  E ( x) 
700

D( x) 
2
800
, (50)

Number of occurence
600

Number of occurence
Number of occurence

700
i
500
500
400
600 M i 1
300 400
300
200
200
where M represent the overall pixels of the image. In the test,
100 100

Pixel value
0 Pixel value
0
1000 pixel pairs were randomly selected for the original Lena
Pixel value
0 50 100
Pixel value
150 200 250 0 50 100
Pixel value
150 200 250 0 50 100
Pixel value
150 200 250 and Fig. 6(b), and the correlation coefficients of all directions
(a1) (a2) (a3) were obtained by the above formulas. The correlations of
400 400 400
original and encrypted Lena image at different directions are
shown in Fig.9. The different pixel values of the original
Number of occurence
Number of occurence

Number of occurence
300 300 300
image are distributed on the diagonal, which indicates that it
200 200 200
has extremely correlation between adjacent pixels of original
100 100 100 image. We can see that all the pixels of the encrypted image
0 Pixel value
0 Pixel value
0 Pixel value
are distributed in the entire plane, which demonstrates that
0 50 100
Pixel value
150 200 250 0 50 100
Pixel value
150 200 250 0 50 100
Pixel value
150 200 250
almost not have correlation between different pixels of
(b1) (b2) (b3) encrypted image.
FIGURE.8 Histogram of image, (a1) histogram of original image R, (a2) 250 250 250

pixel value on (x,y+1)


Pixel value on (x,y+1)
histogram of original image G, (a3) histogram of original image B, (b1)

pixel value on (x,y+1)


200 200 200

histogram of encrypted image R, (b2) histogram of encrypted image G, 150 150 150

(b3) histogram of encrypted image B. 100 100 100

50 50 50
2) CORRELATION COEFFICIENT ANALYSIS
Image correlation means that an important statistical 0
0 50 100 150 200
pixel value on (x, y)
250
0
0 50 100 150
pixel value on (x, y)
200 250
0
0 50 100 150 200
pixel value on (x, y)
250

characteristic of image, the encryption algorithm can be (a1) (a2) (a3)


cracked through statistical analysis. Therefore, in this paper, 250 250 250
pixel gray value on (x, y+1)

pixel gray value on (x, y+1)

pixel gray value on (x, y+1)


the correlation coefficient of different pixels is used to 200 200 200
measure the ability to reduce correlation of the algorithm. If
150 150 150
the correlation coefficient of encrypted image is smaller than
100 100 100
original image, the encryption performance of algorithm is
50 50 50
good and the ability of anti-attack cracking is strong.
0 0 0
Correlation coefficients of different pixels are calculated by 0 50 100 150
pixel gray value on (x, y)
200 250 0 50 100 150 200
pixel gray value on (x, y)
250 0 50 100 150
pixel gray value on (x, y)
200 250

[49] (b1) (b2) (b3)


FIGURE.9 Correlation of images, (a1) correlation of original Lena image
cov( x, y )
rxy  , (47) at horizontal, (a2) correlation of original Lena image at vertical, (a3)
D( x) D( y ) correlation of original Lena image at diagonal, (b1) correlation of
encrypted Lena image at horizontal, (b2) correlation of encrypted Lena
where cov (x, y) represent the covariance of x and y, D(x) image at vertical, (b3) correlation of original Lena image at diagonal
and D(y) means that variance of x and y. According to
mathematical expectation E(x), cov (x, y), D(x) and D(y) are Correlation coefficient values of original and encrypted
calculated by Lena image at different directions are listed in Table. 5.
Correlation coefficient values between pixels of R, G and B
cov( x, y)  E  x  E ( x) y  E ( y) , (48) components at different directions are shown in Table 6.
We can see that correlation coefficient values of original
1 M image are about 1, the encrypted image are almost 0, which
E ( x) 
M
x ,
i 1
i (49) shows that the new algorithm can effectively reduce
correlation between adjacent pixels of original image.
Compared with other algorithms, the proposed color image
compression and encryption scheme has higher security.
TABLE.5
CORRELATION COEFFICIENT OF DIFFERENT DIRECTIONS
Directions Original image Our algorithm Ref.[4] Ref.[5] Ref.[6]
Horizontal 0.9824 0.0016 0.1257 0.0024 0.0044
Vertical 0.9632 0.0015 0.0581 0.0580 0.0034
Diagonal 0.9484 -0.0017 0.0504 0.0170 0.0020

TABLE.6
CORRELATION COEFFICIENT OF DIFFERENT DIRECTION FOR R, G AND B

11 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (March 2019)

Channels Directions Original image Our algorithm Ref.[9] Ref.[19]


Horizontal 0.9774 -0.0020 -0.0206 0.0024
R Vertical 0.9553 -0.0013 -0.0116 0.0010
Diagonal 0.9324 -0.0059 -0.0097 -0.0147
Horizontal 0.9706 -0.0022 -0.0005 -0.0056
G Vertical 0.9443 -0.0041 0.0002 -0.0037
Diagonal 0.9194 0.0014 0.0189 -0.0295
Horizontal 0.9569 0.0069 0.0016 -0.0078
B Vertical 0.9721 0.0059 0.0133 0.0031
Diagonal 0.9015 0.0035 -0.0123 -0.0247

TABLE.8
INFORMATION ENTROPY OF DIFFERENT ENCRYPTION ALGORITHM
Algorithms Our algorithm Ref.[10] Ref.[11] Ref.[12] Ref.[13]
Lena 7.9985 7.9973 7.9972 7.9879 7.9888
image, if C(i, j) and C1(i, j) are not equal, then D(i, j) is 1,
D. INFORMATION ENTROPY otherwise D(i, j) is 0.
The information entropy is a particularly useful measure to Table.9
test the randomness of image information, and it is calculated MEAN NPCRS AND UACIS OF ENCRYPTED LENA IMAGES
Algorithms NPCR (%) UACI (%)
by
Channel R G B R G B
M Our 99.60 99.61 99.61 33.45 33.43 33.41
H   p(i ) log 2 p( j ) , (51)
Ref [4]
Ref [6]
99.42
99.59
99.60
99.22
99.54
98.85
27.78
33.48
27.66
33.46
24.94
33.27
j 0
In our experiments, we only change the lowest bit of one
where M means that the gray scale of the image, and p(j) random pixel of the original image, and carry out the test
represent the probability of gray value j occurrence. For for 10 times with one round of encryption to obtain the
L=256 gray image, the theoretical value of information average NPCRs and UACIs as listed in Table.9. The results
entropy H is 8 [27]. The information entropy values of shows that the mean NPCRs and UACIs of our algorithm
encrypted Lena image for R, G and B in Table 7. Table 8 are over 99.6% and 33.3% respectively only through one
listed the information entropy values of different encryption round of encryption, which shows that the algorithm can
algorithms. From the Table 7 and 8 we can see that new prevent the differential attack.
algorithm has more randomness of image information.
TABLE.7 F. ROBUSTNESS ANALYSIS
INFORMATION ENTROPY OF R, G AND B The robustness is an important measure to evaluate
Images R G B algorithm security. Image robustness refers to that the
Lena 7.9948 7.9958 7.9950 image still has certain fidelity after experiencing various
Ref.[7] 7.9729 7.9744 7.9705 signal processing or various attacks. In this test, robustness
Ref.[8] 7.9732 7.9750 7.9715 is analyzed by cropping attack, rotation attack.
Ref.[9] 7.9914 7.9907 7.9907
1) CROPPING ATTACK
Ref.[19] 7.9893 7.9898 7.9894
To test the algorithm resist cropping attack, we make the
encrypted Lena image with six different data losses as
E. DIFFERENTIAL ATTACK ANALYSIS shown in Fig. 10 (a), (b), (c), (d), (e) and (f). The
Researchers usually use pixel count change rate (NPCR) corresponding decrypted images are presented in Fig. 10 (g),
and average intensity change rate (UACI) as two criteria to (h), (i), (j), (k) and (l), respectively. The Fig. 10 shows that
measure whether the method can resist differential attack. even though the encrypted images are lost, the main
NPCR and UACI calculation as follows: information of image can be covered. Therefore, our
 D(i, j ) algorithm would resist data loss attack in different degree.
NPCR  100% , (52)
i, j

L
1 C (i, j )  C1 (i, j )
UACI  
L i, j 256
100% , (53)

here, C is encrypted image pixel value before original


image changed, and C1is encrypted image pixel values after
original image changed, L is the number of all pixels in an (a) (b) (c)

11 VOLUME XX, 2019

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (March 2019)

indicates that the pseudo-random sequence generated by


fractional-order memristor chaotic circuit has better
randomness. Moreover, it has some advantages of image
information encryption. Based on this chaotic system, a
novel color image compression and encryption algorithm is
proposed. The compression ratio research results show that
(d) (e) (f) the algorithm has stronger compression and reconstruction
effects. The security performance analysis illustrates that the
designed algorithm can resist various attack. Therefore, we
presented image compression and encryption scheme can
effectively compress and encrypt image, which provides
experimental basis and theoretical guidance for safe
transmission of image information. In the later work, we will
propose more encryption and CS algorithms and
(g) (h) (i) continuously improve security.

REFERENCES
[1] Li. C, Sprott. J. C, and Xing. H, “Constructing chaotic systems with
conditional symmetry,” Nonlinear Dynamics., vol. 87, no. 2, pp.
1351-1358, Jan. 2017.
[2] Li. C, Sprott. J. C, and Mei Y, “An infinite 2-D lattice of strange
attractors,” Nonlinear Dynamics., vol. 89, no. 4, pp. 2629-2639, Sep.
2017.
[3] Cao. C, Sun. K, and Liu. W, “A novel bit-level image encryption
(j) (k) (l) algorithm based on 2D-LICM hyperchaotic map,” Signal Processing.,
vol. 432, pp. 122-133, Feb. 2018.
FIGURE.10 Cropping attack analysis results, (a) 1/2 data loss of [4] Huang. C. K, and Nien. H. H, “Multi chaotic systems based pixel
Figure.6(a2) , (b) 1/4 data loss of Figure.6(a2), (c) 1/8 data loss of shuffle for image encryption,” Optics Communications., vol. 282, no.
Figure.6(a2), (d) 1/16 data loss of Figure.6(a2), (e) 1/32 data loss of 11, pp. 2123-2127, 2009.
[5] Liu. H, Wang. X, and Kadir. A, “Color image encryption using
Figure.6(a2), (f) 1/32 data loss of Figure.6(a2), (g) decrypted image of 1/2
Choquet fuzzy integral and hyper chaotic system,” Optik -
data loss, (h) decrypted image of 1/4 data loss, (i) decrypted image of International Journal for Light and Electron Optics., vol. no. 18, pp.
1/8 data loss, (j) decrypted image of 1/16 data loss, (k) decrypted image 3527-3533, Sep. 2013.
[6] X. Wei, L. Guo, Q. Zhang, J. Zhang, and S. Lian, “A novel color
of 1/32 data loss, (l) decrypted image of 1/64 data loss
image encryption algorithm based on DNA sequence operation and
2) ROTATION ATTACK hyper-chaotic system,” J. Syst. Softw., vol. 124, no. 23, pp. 6276-6281,
Dec. 2013.
The rotation attack is a typical geometric attack. If the [7] A. Kadir, A. Hamdulla, and W. Guo, “Color image encryption using
attacker made slight geometric transformation on the skew tent map and hyper chaotic system of 6th-order CNN,” Optik.,
encrypted image, the pixel position in the image is almost vol. 125, no. 5, pp. 1671–1675, Mar. 2014.
completely changed, which make the image owner unable [8] R. Rhouma, S. Meherzi, and S. Belghith, “OCML-based colour image
encryption,” Chaos Solitons Fractals., vol. 40, no. 1, pp. 309–318,
to get the properly encrypted image. For this test, the cipher Apr. 2009.
image is rotated with 2°, 18°and 25°, and the test results as [9] Wu. X, Wang. D, Kurths. J, and Kan. H, “A novel lossless color
show in Fig. 11. The test results indicate that the algorithm image encryption scheme using 2D DWT and 6D hyperchaotic
system,” Information Sciences., vol. 349–350, pp. 137-153, Jul. 2016.
can resist rotation attack. [10] L. Xu, X. Gou, Z. Li, and J. Li, “A novel chaotic image encryption
algorithm using block scrambling and dynamic index based
diffusion,” Opt. Lasers Eng. vol. 91, pp. 41–52, Apr. 2017.
[11] Y. Li, C. Wang, and H. Chen, “A hyper-chaos-based image
encryption algorithm using pixel-level permutation and bit-level
permutation,” Opt. Lasers Eng., vol. 90, pp. 238–246, Mar. 2017.
[12] S.M. Ismail, L.A. Said, A.A. Rezk, A.G. Radwan, A.H. Madian, M.F.
Abu-ElYazeed, and A. M, “Soliman, Biomedical image encryption
based on double-humped and fractional logistic maps,” in: 2017 6th
International Conference on Modern Circuits and Systems
(a) (b) (c) Technologies (MOCAST), IEEE, May, pp. 1–4, 2017.
FIGURE.11 Rotation attack results, (a) recovered image of rotation 2°, (b) [13] J.F. Zhao, S.Y. Wang, L.T. Zhang, and X.Y. Wang, “Image
encryption algorithm based on a novel improper fractional-order
recovered image of rotation 18°, (c) recovered image of rotation 25°
attractor and a wavelet function map,” J. Electr. Comput. Eng., pp. 1-
10, 2017.
VII. CONCLUSION [14] Norouzi. B, and Mirzakuchaki. S, “A fast color image encryption
In this paper, a fractional-order memristor chaotic circuit is algorithm based on hyper-chaotic systems,” Nonlinear Dynamics., vol.
78, no. 2, pp.995-1015, Oct. 2014.
obtained. The dynamic behaviors analysis shows that the [15] Seyedzadeh. S. M, Norouzi. B, Mosavi. M. R, and Mirzakuchaki. S,
fractional-order chaotic system represented better sensitivity “A novel color image encryption algorithm based on spatial
of initial values and parameters. The randomness test

VOLUME XX, 2019 14

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (March 2019)

permutation and quantum chaotic map,” Nonlinear Dynamics., vol. 8, [38] Donoho. D. L. “Compressed sensing,” IEEE Transactions on
no. 1-2, pp. 511-529, Jul. 2015. Information Theory., vol. 52, no. 4, pp. 1289-1306, 2006.
[16] Wang. X, and Zhang. H. L, “A novel image encryption algorithm [39] Tong. X. J, Zhang. M, Wang. Z, and Ma. J, “A joint color image
based on genetic recombination and hyper-chaotic systems,” encryption and compression scheme based on hyper-chaotic system,”
Nonlinear Dynamics., vol. 83, no. 1-2, pp. 333-346, Jan. 2016. Nonlinear Dynamics., vol. 84, no. 4, pp. 2333-2356, (2016).
[17] Wu. X, Wang. K, Wang. X, and Kan. H, “Lossless chaotic color [40] Zhu. H, Zhao. C, and Zhang. X, “A novel image encryption–
image cryptosystem based on DNA encryption and entropy,” compression scheme using hyper-chaos and Chinese remainder
Nonlinear Dynamics., vol. 90, no. 2, pp. 855-875, Oct. 2017. theorem,” Signal Processing Image Communication., vol. 28, no. 6,
[18] Sahari. M. L, and Boukemara. I, “A pseudo-random numbers pp. 670-680, 2013.
generator based on a novel 3D chaotic map with an application to [41] Zhang. D, Liao. X, Yang. B, and Zhang. Y, “A fast and efficient
color image encryption,” Nonlinear Dynamics., vol. 94, no. 1, pp. approach to color-image encryption based on compressive sensing and
723-744, Oct. 2018. fractional Fourier transform,” Multimedia Tools & Applications., vol.
[19] Chai. X, Fu. X, Gan. Z, Lu. Y, and Chen. Y, “A color image 77, no. 2, pp. 2191-2208, Jan. 2018.
cryptosystem based on dynamic DNA encryption and chaos,” Signal [42] Chai. X, Gan. Z, Chen. Y and Zhang. Y, “A visually secure image
Processing, vol. 155, pp. 44-62, 2019. encryption scheme based on compressive sensing,” Signal Processing,
[20] Gan. Z, Chai. X, Zhang. M and Lu. Y, “A double color image vol. 134, pp. 35-51, 2016.
encryption scheme based on three-dimensional brownian motion,” [43] Candès, and Emmanuel. J, “The restricted isometry property and its
Multimedia Tools and Applications, 2018. implications for compressed sensing,” Comptes rendus –
[21] Chai. X. L, Gan. Z. H, Lu. Y, and Zhang. Y. S, “A novel color image Mathématique., vol. 346, no. 9, pp. 589-592, 2008.
encryption algorithm based on genetic recombination and the four- [44] Han. Bao, Tao. Jiang, Kaibin. Chu, Chen, Xu. Q, and Bao. B,.
dimensional memristive hyperchaotic system,” Chinese Physics B, vol. “Memristor-Based Canonical Chua’s Circuit: Extreme Multistability
25, no. 10, pp. 76-88, 2016. in Voltage-Current Domain and Its Controllability in Flux-Charge
[22] Wu. X, Li. Y, and Kurths. J, “A new color image encryption scheme Domain,”complexity., 2018.
using CML and a fractional-order chaotic system,” Plos One, DOI: [45] Rukhin. A. L, Soto. J, Nechvatal. J. R, “SP 800-22 Rev. 1a. A
10.1371/journal.pone.0119660 Statistical Test Suite for Random and Pseudorandom Number
[23] Wang. X, Wang. Q, and Zhang, Y, “A fast image algorithm based on Generators for Cryptographic Applications,” Applied Physics Letters.,
rows and columns switch,” Nonlinear Dynamics., 2015, vol. 79, no. 2, vol. 22, no. 7, pp. 1645-179, 2010.
pp.1141-1149, 2015. [46] Chai. X, Zheng. X, Gan. Z, Han. D, and Yi. C, “An image encryption
[24] Oldham. K. B, and Spanier. J, “The fractional calculus: theory and algorithm based on chaotic system and compressive sensing,” Signal
applications of differentiation and integration to arbitrary order,” Processing., vol. 148, pp.124-144, Jul. 2018.
Dover Publications, Inc, 2002. [47] Chai. X, Gan. Z, Yang. K, Yi. C, and Xiang. L, “An image encryption
[25] Bhalekar. S, “Dynamical analysis of fractional order Uçar prototype algorithm based on the memristive hyperchaotic system, cellular
delayed system,” Signal Image and Video Processing., vol. 6, no. 3, automata and DNA sequence operations,” Image Communication., vol.
pp. 513-519, 2016. 52, pp. 6-19, Mar. 2017.
[26] He. S, Sun. K, and Wang. H, “Complexity Analysis and DSP [48] Liu. W, Sun. K, He. Y, and Yu. Y, “Color Image Encryption Using
Implementation of the Fractional-Order Lorenz Hyperchaotic Three-Dimensional Sine ICMIC Modulation Map and DNA Sequence
System,”Entropy., vol. 17, no. 12, pp. 8299-8311, Dec. 2015. Operations,” International Journal of Bifurcation & Chaos., vol. 27,
[27] Zhang. L, Sun. K, He. S, Wang. H, and Xi. Y, “Solution and dynamics no. 11, 2017.
of a fractional-order 5-D hyperchaotic system with four wings,” [49] Zhu. C, and Sun K, “Cryptanalyzing and Improving a Novel Color
European Physical Journal Plus., Jan. 2017. Image Encryption Algorithm Using RT-Enhanced Chaotic Tent
[28] Ruan. J, Sun. K, Mou. J, Zhang. L, and He. H, “Fractional-order Maps,” IEEE Access., 2018.
simplest memristor-based chaotic circuit with new derivative,”
European Physical Journal Plus., Jan. 2018.
[29] Chua. L. O, and Alexander. G, “The effects of parasitic reactances on
nonlinear networks,” IEEE Trans Circuits Syst., vol. 18, no. 5, pp.
520-532, 1971.
[30] Wu. H, Bao. B, Liu. Z, Quan. X, and Pin. J, “Chaotic and periodic
bursting phenomena in a memristive Wien-bridge oscillator,”
Nonlinear Dynamics., vol. 83, no. 1-2, pp. 893-903, Jan. 2016.
[31] Yu. Q, Bao. B. C, Hu. F. W, Xu. Q, Cheng. M, and Wang. J, “Wien-
bridge chaotic oscillator based on fisrt-order generalized memristor,”
Acta Physica Sinica., vol. 63, no. 24, pp. 240505-240505, 2014.
[32] Chen. M, Li. M, Yu Q, Bao. B, Xu. Q, and Wang. J, “Dynamics of
self-excited attractors and hidden attractors in generalized memristor-
based Chua’s circuit,” Nonlinear Dynamics., vol. 81, no. 1-2, pp. 215-
226, Jul. 2015.
[33] Wu. H, Bao. B, Liu. Z, Xu. Q, and Jiang. P, “Chaotic and periodic
bursting phenomena in a memristive Wien-bridge oscillator,”
Nonlinear Dynamics., vol. 83, no. 1-2,pp. 893-903, Jan. 2016.
[34] Strukov D B, Snider G S, Stewart D R, and R. Stanly. Williams, “The
missing memristor found,” Nature., vol. 453, no. 7191, pp. 80-83,
2008.
[35] Ye. X, Mou. J, Luo. C, and Wang Z, “Dynamics analysis of Wien-
bridge hyperchaotic memristive circuit system,” Nonlinear Dynamics.,
vol. 92, no. 3, pp. 1-11, May. 2018.
[36] Tan. Q, Zeng. Y, and Li. Z, “A simple inductor-free memristive
circuit with three line equilibria,” Nonlinear Dynamics., vol. 94, no. 3,
pp. 1585-1602, Nov. 2018.
[37] Sun. J, Zhao. X, Fang. J, and Wang. Y, “Autonomous memristor
chaotic systems of infinite chaotic attractors and circuitry realization,”
Nonlinear Dynamics., vol. 94, no. 4, pp. 2879-2887, Dec. 2018.

VOLUME XX, 2019 14

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2914722, IEEE Access
Author Name: Preparation of Papers for IEEE Access (March 2019)

FEIFEI YANG received the B.E. degree


Longdong University, Qingyang, China, in 2016.
He is currently pursuing the Ph.D. degree in
control science and engineering at Dalian
polytechnic University, Dalian, China. He
mainly research interest includes chaos theory
and application.

JUN MOU received the B.S. degree, the M.S.


degree and Ph.D. degree in physics and
electronics from Central South University,
Changsha, China. He is currently an associate
professor at the School of Information Science
and Engineering, Dalian Polytechnic University,
China. His mainly research interest includes the
nonlinear system control, secure communication,
power system automation and smart grid
research.

KEHUI SUN received the B.S degree in


industrial automation, and the Ph.D. degree in
control theory and control engineering from
Central South University, Changsha, China, in
1998 and 2005 respectively. He is a professor at
the School of Physics and Electronics, Central
South University, Changsha, China currently. His
main research includes the chaos synchronization
control theory and its application in secure
communication, intelligent instrument
development.

YINGHONG CAO received the B.S. degree in


electronic engineering, and the Ph.D. degree in
signal and information processing from Dalian
University of Technology (DUT), Dalian, China,
in 2003 and 2013 respectively. She is currently a
lecturer at the School of Information Science and
Engineering, Dalian Polytechnic University, Her
research interests include communication signal
processing, speech processing, and Internet of
Things Technology and Application, etc.

JIYU JIN received his PhD degree in


information and communication engineering
from Yeungnam University, Gyeongsan, Rep. of
Korea, in 2007. He is currently an associate
professor at the School of Information Science
and Engineering, Dalian Polytechnic University,
China. His research interests include
wireless/mobile communication systems and
Internet of things.

VOLUME XX, 2019 14

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

You might also like