Accepted Manuscript

Image Encryption Using 2D Hénon-Sine Map and DNA Approach

Jiahui Wu, Xiaofeng Liao, Bo Yang

PII: S0165-1684(18)30205-6
DOI: 10.1016/j.sigpro.2018.06.008
Reference: SIGPRO 6845

To appear in: Signal Processing

Received date: 12 December 2017

Revised date: 8 April 2018
Accepted date: 6 June 2018

Please cite this article as: Jiahui Wu, Xiaofeng Liao, Bo Yang, Image Encryption Using 2D Hénon-Sine
Map and DNA Approach, Signal Processing (2018), doi: 10.1016/j.sigpro.2018.06.008

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Highlights

• We propose a novel 2D chaotic map named 2D-HSM, which is generated by connecting Henon map and
Sine map. Performance evaluations including phase diagram, bifurcation diagram and Lyapunov exponent
spectrum illustrate that the new map has better ergodicity, wider chaotic region, and is more unpredictable
than other 2D chaotic maps.

• We analyse the shortcomings of an existing image encryption model (SPDT), and find that the model cannot
resist statistical attack, and more importantly, the encryption model is irreversible, which illustrates that the

T
model cannot even guarantee a normal intercommunication.

IP
• To resist statistical attack, equivalent key streams of an encryption model needs to be random and homoge-

CR
neous. Using the proposed 2D chaotic map, we design a new chaos-based encryption algorithm with equi-
librium key streams, and a DNA encryption is designed in the encryption model to enhance the encryption
efficiency.

US
• We verify the security of the proposed scheme in experimental simulation and comparisons from four aspects:
the exhaustive attack, the statistical attack, the differential attack and the noise attack. Simulation results show
AN
that the proposed algorithm has better security.

• We provide explanations of experimental phenomena in good key sensitivity, resisting differential attack and
robustness against noise for someone who believe the three performances seem to contradict each other and
M

cannot exist in an encryption scheme at the same time.

ED
PT
CE
AC

1
 ACCEPTED MANUSCRIPT

Image Encryption Using 2D Hénon-Sine Map and DNA Approach

Jiahui Wu, Xiaofeng Liao∗, Bo Yang

Chongqing Key Laboratory of Nonlinear Circuits and Intelligent Information Processing,
College of Electronic and Information Engineering, Southwest University, Chongqing 400715, China

T
IP
Abstract: Chaotic systems have been widely applied in digital image encryption due to their complex properties
such as ergodicity, pseudo randomness and extreme sensitivity to their initial values and parameters. This paper

CR
first proposes a two-dimensional Hénon-Sine map (2D-HSM). The new map possesses better ergodicity and pseudo
randomness, and its parameters have a wider chaotic range, compared with many existing chaotic systems. Then
a DNA encoding and a DNA exclusive-or (XOR) operation rule are defined because DNA approach applying in

US
image encryption can greatly improve the efficiency of image permutation and diffusion. Furthermore, a novel
image encryption scheme whose image pixels are diffused by the DNA approach and permutated by 2D-HSM, is
AN
proposed to protect image content while an image is transfered over the Internet. Some experimental analyses such
as statistical attack analysis, differential attack analysis, exhaustive attack analysis, robustness against noise and
computational complexity have been applied to measure the new scheme, and the experimental results illustrate
M

the scheme possesses better encryption performances than that of other references, and therefore, is secure in real-
world communication.
ED

Keywords: Image encryption; DNA sequence; 2D Hénon-Sine map; Permutation; Diffusion.

1 Introduction
PT

The rapid development of digital technology and the popularity of the Internet have brought great convenience
to people’s work and life. Digital media as the carrier of works such as books, music, images and videos, etc,
CE

have greatly enriched people’s lives due to their easy access, convenient copying, quick spread and other advan-
tages. However, as we can see, some of the malicious behaviors aiming to intercept useful information using the
AC

characteristics of network openness and sharing, have seriously damaged the interests of communication parties.
Therefore, it is urgent to develop the technology of secure information communication.
Digital images with visual visibility are an important form of media data and have been widely spread over
the Internet. Therefore, image encryption for securely transmit digital image over the Internet has become a hot
technology. Due to some special properties of images such as large data capacity, strong correlation of pixel points
and high redundancy, the developed data encryption algorithms such as DES and AES [1, 2] which take a lot of time
∗
E-mail address of corresponding author:xfliao@swu.edu.cn

2
 ACCEPTED MANUSCRIPT

to calculate, are not suitable for the secure transmission of real-time images, and thus a large number of algorithms
dedicated to image encryption have been designed based on chaotic map [3–14], deoxyribonucleic acid (DNA)
sequence [12, 13], lattice cipher [15, 16], cellular automata [17, 18], transform domain [19, 20], magic cube [29]
and so on.
Chaotic system possesses the prominent features, including extremely sensitive dependence on initial conditions
and system parameters, ergodicity and random-like behaviors, which is very suitable for image encryption. Fridrich
first proposed chaos-based image encryption scheme in 1997 [3]. Subsequently, chaos-based image encryption
algorithms have growingly developed [4–14]. Image encryption in [4] is a asymmetric scheme using elliptic curve

T
ElGamal. The scheme is secure, but it has drawbacks in resisting noise and cropping attack, and the asymmetric

IP
encryption scheme has the same defect in long time encryption. Therefore, the application of symmetric image
encryption scheme is more extensive. In [7, 8], two symmetric models which both shuffle image with a random

CR
matrix generated by Logistic map, and then diffuse the shuffled image by XOR operation determined by another
chaotic map are proposed. But, the scheme in [7] must be performed multiple times to get good performance, and
the scheme in [8] is vulnerable to differential attack. In [9–11], the authors propose their respective new chaotic

US
map by compounding multiple existing chaotic systems, respectively, and then propose their own symmetric image
cipher based on the new chaotic maps. In [12–14], different methods of image encryption based on chaos are
AN
proposed. The authors in [14] integrate the advantages of both chaos and optics to design a new encryption scheme,
but its decipher image has a certain distortion.
Although various chaos-based image encryption algorithms have been proposed by many researchers, some of
M

them are proven to have insufficient security [23–28]. Image encryption schemes based on one-dimensional chaotic
systems [5–7] have simple chaotic orbits and can be cracked by brute-force attack [29]. Although high-dimensional
chaotic systems possess complicated chaotic behaviors and are difficult to determine their chaotic orbits, there still
ED

exist weaknesses, such as complicated performance analyses and high implementation consumptions [32]. In view
of this, this paper proposes a novel 2D chaotic map named 2D-HSM, which is generated by connecting Hénon map
PT

and Sine map. Performance evaluations including phase diagram, bifurcation diagram and Lyapunov exponent
spectrum illustrate that the 2D-HSM has better ergodicity, wider chaotic region, and is more unpredictable than
other 2D chaotic maps.
CE

In recent years, DNA-based image encryption schemes have received increasing concerns by many researchers
because DNA technology possesses high parallelism, massive storage and low-power consumption, and DNA
AC

computing has also been conducted at the same time [12]. A. Gehani put forward a DNA-based one-time-pad
system [21], which took advantage of DNA’s small volume and large storage to facilitate the storage of a large
number of keys. However, at present, the concrete realization of the large-scale cipher book of one-time-pad based
on DNA requires complex biological experiments, and therefore, it needs to be deserved further studies. The
encryption method based on DNA encoding has been proposed by many scholars.
There are generally two methods for image encryption based on DNA encoding: firstly, encode all the plain-
image pixels into DNA sequence according to DNA theory, that is, decompose each of the image-pixel 8-bit gray

3
 ACCEPTED MANUSCRIPT

value into four DNA codes so as to reduce the time complexity of permutation and diffusion in image encryption.
Secondly, according to a defined DNA operation rules (addition, subtraction or XOR operation), the encoded plain
image is operated to form a cipher image. Zhang et al. propose an encryption algorithm based on DNA plus
operation and chaos theory [22]. They encode chaotic sequence and plain image into DNA sequence, and then add
the two sequences with DNA addition rule to form an incomprehensible sequence which is the cipher sequence.
Although the algorithm has good encryption effect in the performance of information entropy and adjacent pixel
correlation, it can not resist differential attack. SPDT proposed in [35] is also an example of image encryption
based on DNA operation. The general idea of SPDT is utilizing key streams generated by chaotic map to determine

T
the DNA encoding/decoding rules first, and then encodes/decodes pixel values according to DNA rules and an extra

IP
XOR operation, after that, the diffused image is obtained and further permutated. This encryption scheme makes
encryption process faster than other cryptosystems, and can resist common attacks. However, it has the following

CR
vital defects which have been neglected: in key-streams generation phase, not all secret keys can generate proper
initial values which meet the chaotic conditions. In the diffusion phase, the rules generated by key streams are not
uniform, so that attackers can easily obtain some useful information of plain image by statistical analyses. In the

US
permutation phase, a new permutation map is proposed and can effectively resist common attacks, but its inverse-
permutation method does not exist although the permutation map is a one-to-one mapping. To overcome all of the
AN
aforementioned defects, a new encryption scheme based on 2D-HSM and DNA is proposed. In diffusion phase, the
new scheme generates an uniform DNA rules and embeds a statistical characteristic of plain image into diffusion
step so as to resist common attacks (The known plaintext attack and the chosen plaintext/ciphertext attack). In
M

permutation phase, the relation between permutation map and image pixel value of SPDT is canceled so that the
map method is determined solely by secret keys, and can be reversed with secret keys. Therefore, receiver can
decrypt the scheme only when he has secret keys.
ED

The rest of this paper is arranged as follows: Section 2 introduces the existing chaotic maps and proposes a new
chaotic map named 2D-HSM. Section 3 further proposes an image encryption and its corresponding decryption
PT

schemes based on DNA and 2D-HSM. Section 4 analyzes the security performance including statistical attack
analysis, differential attack analysis, exhaustive attack analysis, robustness against noise attack and computational
complexity, and compares the novel scheme with other three schemes, and gives an explanation to eliminate the
CE

possible confusion about our experimental phenomena in key sensitivity, differential attack and noise attack. Fi-
nally, Section 5 presents a brief conclusions.
AC

2 2D Hénon-Sine map

This section constitutes Hénon map and Sine map to generate a 2D Hénon Sine map (2D-HSM) and deliberates
its chaotic behaviors.

4
 ACCEPTED MANUSCRIPT

2.1 Mathematic definition

Sine map is a 1D chaotic map and is defined by the following equation.

xn+1 = µ sin(πxn ), (1)

where µ ∈ [0, 1]. When µ ∈ [0.87, 1], the map is chaotic.

Hénon map is a 2D discrete chaotic map, which is defined as


T


 2
 xn+1 = 1 − axn + yn ,

 (2)

IP



yn+1 = bxn ,

CR
when parameter b = 0.3 and a ∈ [1.06, 1.22] ∪ [1.27, 1.29] ∪ [1.31, 1.42], Hénon map is chaotic [32]. Trajectory
of a dynamical system is to present the path of system’s outputs. Fig. 1 (a-b) show the trajectories of Sine map and
Hénon map with the same initial values (0.3, 0.4). These two figures show that the trajectories of both the above

US
maps are simple and have something determined, thus both of the maps can be easily predicted [29]. To address
this defect, we propose a complicated and unpredicted chaotic map. The new map named 2D-HSM is defined by
the following equation,
AN



 2
 xn+1 = (1 − a sin (xn ) + yn ) mod 1,


 (3)


yn+1 = bxn mod 1,
M

where the parameters of a and b are both expanded to (−∞, +∞).

2.2 Performance evaluation

ED

The proposed 2D-HSM possesses good chaotic behaviors compared with the original two maps and some other
maps. Here, the trajectory, bifurcation and Lyapunov exponent are used to evaluate the map’s chaotic performance.
PT

2.2.1 Trajectory
CE

We set initial value of 2D-HSM as the same as Sine map and Hénon map. Fig. 1 (c) shows the trajectories of
2D-HSM. We can see that the 2D-HSM distributes in the whole range of phase plane and has larger distribution
area compared with phase diagrams of Sine and Hénon map shown in Fig. 1 (a,b). Therefore, 2D-HSM has better
AC

ergodicity and more random outputs, compared with Sine and Hénon maps.

2.2.2 Lyapunov exponent

An important feature of chaos is that a tiny uncertainty of the initial state will result in exponentially expanded
outputs at a high speed. In a non-chaotic system, the trajectories close to each other either converge exponentially
or diverge at an exponential rate. The rate of convergence or divergence of system’s trajectories can be described by

5
 ACCEPTED MANUSCRIPT

the Lyapunov exponent (LE) λ [32, 33]. Lyapunov exponents of 2D-HSM are calculated according to the complete
mathematical definition in [34]. A positive LE represents that even a slight change in initial state, the difference of
outputs is completely different eventually. Thus, dynamical system is chaotic when λ > 0. Fig. 1 (d-f) and Fig. 1
(g-i) present the bifurcation diagram and LE spectrum of Sine map, Hénon map and HSM with their corresponding
parameters, respectively. From these subgraphs, we know that 2D-HSM has chaotic behavior when approximately
a ∈ (−∞, −0.71] ∪ [0.71, ∞), b = 0.7. In fact, as shown in Fig. 3, when b < [−1, 1], a ∈ R, the largest Lyapunov
exponent of 2D-HSM is greater than 0. Compared with LE value of other 2D chaotic maps [30–32] shown in
Fig. 2 (a,b,c), the 2D-HSM has wider chaotic range, greater LE values and therefore, possesses more complicated

T
trajectories and its outputs are more unpredictable.

IP
1 0.4 1

CR
0.9
0.9 0.3
0.8
0.2
0.8 0.7
0.1
0.6
0.7
x y 0
y 0.5
n+1
0.6

0.5

0.4
−0.1

−0.2

−0.3
US 0.4

0.3

0.2

0.1
AN
−0.4 0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 −1.5 −1 −0.5 0 0.5 1 1.5 0 0.2 0.4 0.6 0.8 1
xn x x

(a) Phase diagram of Sine map (b) Phase diagram of Hénon map (c) Phase diagram of 2D-HSM

1 1.2 1
M

0.9 0.9
1
0.8 0.8

0.7 0.8 0.7

0.6 0.6
0.6
ED

x 0.5 x x 0.5
0.4
0.4 0.4

0.3 0.2 0.3

0.2 0.2
0
0.1 0.1
PT

0 −0.2 0
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 −10 −5 0 5 10
µ a a

(d) Bifurcation diagram of Sine map (e) Bifurcation diagram of Hénon map (f) Bifurcation diagram of 2D-HSM
CE

1 1 3

0.5 λ1 2
0.5
λ2
0
1

−0.5 0
AC

0
−1
LE LE−0.5 LE−1 λ1

−1.5 λ2
−2
−2 −1

−3
−2.5
−1.5
−3 −4

−3.5 −2 −5
0 0.2 0.4 0.6 0.8 1 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 −10 −5 0 5 10
µ a a

(g) Lyapunov exponent diagram of Sine map (h) Lyapunov exponent diagram of Hénon map (i) Lyapunov exponent diagram of 2D-HSM
Fig. 1 Sine map, Hénon map and 2D-HSM.

6
 ACCEPTED MANUSCRIPT

T
IP
(a) 2D Logistic map (b) 2D SLMM map (c) 2D LASM map
Fig. 2 Lyapunov exponents of other 2D chaotic maps.

CR
US
AN

(a) LE with changing a, b (b) Top view of (a) (c) Left view of (a)
M

Fig. 3 Lyapunov exponents of 2D-HSM.

3 2D-HSM-based image encryption algorithm

ED

This section designs a novel image encryption algorithm which combines the above proposed 2D-HSM and a
DNA approach.
PT

3.1 Preliminary tools

CE

Deoxyribonucleic Acid (DNA): a DNA sequence has four nucleic acids, they are the purine A (adenine), G
(guanine) and the pyrimidine T (thymine), C (cytosine) respectively. A always paires with T, while C with G,
which is called Watson-Crick base pairing rules. Table 1 shows the 8 rules of DNA encoding and decoding. For
AC

example, the binary of number 201 is (11001001)2 , then it can be encoded to (ATGC) under Rule 4 in Table 1.
Table 2 shows an XOR operation of nucleic acids which is applied in diffusion phase.

3.2 Encryption

The encryption process includes the diffusion by a DNA approach and the permutation by a generated sequence.
2D-HSM is used to control the DNA rules, DNA operation and the permutation sequence. The flowchart of the

7
 ACCEPTED MANUSCRIPT

Table 1. Encoding and decoding rules of DNA sequence.

A T C G
Rule 1 00 11 10 01
Rule 2 00 11 01 10
Rule 3 11 00 10 01
Rule 4 11 00 01 10
Rule 5 10 01 00 11
Rule 6 01 10 00 11
Rule 7 10 01 11 00
Rule 8 01 10 11 00

T
Table 2. XOR operation of DNA sequence.

IP
XOR A T C G
A A T C G

CR
T T A G C
C C G A T
G G C T A

US
encryption process is shown in Fig. 4 while a simple example of using the encryption scheme to encrypt a 3 × 3
image is shown in Fig. 5. The detailed steps of the encryption algorithm are presented in the following.
AN
Step 1: Choose secret keys {a1 , b1 , x10 , y01 , a2 , b2 , x20 , y02 } which are the parameters and initial values of 2D-HSM.
Step 2: Obtain chaotic key-streams x1 , y1 and x2 , y2 using Eq. (3) with control parameters and initial values
a1 , b1 , x10 , y01 and a2 , b2 , x20 , y02 , respectively.
M

Step 3: Read a plain image Im×n , and reshape I into one-dimensional array P of length mn.
Step 4: Diffuse sequence P using Algorithm 1 to obtain a diffused sequence New P.
ED

Step 5: Permutate New P with Algorithm 2 to obtain the final encrypted sequence C.
Step 6: Reshape C into Ic of size m × n, Ic is the encrypted image.
The above scheme is designed mainly for encrypting gray image. For RGB color image, the scheme first divides
PT

it into R, G and B components, then, encrypts each component of the image and integrates the three components to
form the final cipher image.
CE

3.3 Decryption

The detailed decryption process is presented in Algorithm 3. In this decryption algorithm, the receiver can
AC

decrypt the cipher image only when he knows the secret keys {a1 , b1 , x10 , y01 , a2 , b2 , x20 , y02 }. There is only one pixel
value (the first one) of the decipher image different from the original plain image. Fig. 6 (a-c) show the original
plain image, the corresponding cipher image encrypted by the proposed scheme and its corresponding decipher
image, respectively. PSNR represented peak signal-to-noise ratio is defined as Eq. (4). Plain image shown as Fig. 6
(a) is seen as a complete signal image, then PSNR value of its cipher image shown as Fig. 6 (b) is 9.140546, and
PSNR value of the corresponding decipher image shown as Fig. 6 (c) is 60.069210, which illustrate that cipher
image of the proposed scheme is similar to noisy image, and the decipher image restores the original image almost

8
 ACCEPTED MANUSCRIPT

R Modular and x1 y1 x1
Rounding operation x2 y2
Loc
Rz Ry Rx

DNAR

T
New_Pi New_Pi

IP
DNA_Pi

CR
Fig. 4 The proposed image encryption process.

undistorted. US
AN
2
Imax
PS NR =10 log10 ,
MS E
m n
1 XX (4)
MS E = (P(i, j) − I(i, j))2 ,
mn i=1 j=1
M

where P and I represent two images which both have dimensions m × n, and i, j locate the pixel position.
ED

Algorithm 1 The proposed diffusion algorithm

Input: Pixel value Pi (the ith pixel value of P), i = 1, 2, · · · , mn, and key-streams x1 , y1 , x2 , y2 .
Output: The diffused pixel value New Pi .
PT

P
1: temp = i (Pi ) mod 256.
2: R x ← Round(x1i × 1010 ) mod 8 + 1;
Ry ← Round(yi1 × 1010 ) mod 8 + 1;
CE

Rz ← Round(x2i × 1010 ) mod 8 + 1.

3: R ← Round(yi2 × 1010 ) mod 256.
4: DNAR ← Encode R with rule Rz .
5: DNA Pi ← Encode Pi with rule Ry .
AC

6: New Pi ← DNA Pi ⊕ DNAR .

7: New Pi ← Decode New Pi with rule R x .
8: New Pi ← New Pi ⊕ temp.
9: temp ← New Pi .

9
 ACCEPTED MANUSCRIPT

Algorithm 2 The proposed permutation algorithm

Input: Pixel value Pi and the diffused pixel value New Pi .
Output: A permutated sequence C.
1: S = Round(x1 × 1010 ) mod mn + 1.
2: Remove the redundant elements in sequence S to obtain Loc
which is not repeated and contains all of the data from 1 to mn.
3: Move the ith element of P to Loc(i)th as follows:
4: FOR i = 1 : mn
5: C(Loc(i)) = P(i).
6: END FOR

T
Algorithm 3 The proposed decryption algorithm

IP
Input: The cipher image Ic of size m × n, the secret keys a1 , b1 , x10 , y01 , a2 , b2 , x20 , y02 .
Output: The decrypted image Id .

CR
1: A vector C ← Reshape Ic .
2: Obtain chaotic key-streams x1 , y1 and x2 , y2 using Eq. (3)
with control parameters and initial values a1 , b1 , x10 , y01 and a2 , b2 , x20 , y02 , respectively.
3: R x ← Round(x1 × 1010 ) mod 8 + 1, Ry ← Round(y1 × 1010 ) mod 8 + 1,

4: S ← Round(x1 × 1010 ) mod mn.

5: Loc ← Remove the redundant elements in S .
6: DNAR ← Encode R with Rule Rz .
US
Rz ← Round(x2 × 1010 ) mod 8 + 1, R ← Round(y2 × 1010 ) mod 256.
AN
7: Dd (1) ← C(1).
8: FOR i from 2 to mn
temp ← C(i − 1)
Dd (i) ← C(i) ⊕ temp
M

Dd (i) ← Encode Dd (i) with Rule R x (i)

Dd (i) ← Dd (i) ⊕ DNAR (i)
Dd (i) ← Decode Dd (i) with Rule Ry (i)
ED

9: END FOR
10: D ← Inverse permutation Dd according to Loc.
11: Id ← Reshape D to a matrix of size m × n.
PT

4 Experimental simulation results

CE

In this section, various experiments are analyzed and compared to confirm the security of the proposed scheme.
All the results are run in a Matlab 8.3.0.532 (R2014a) platform on a desktop computer with an Intel Core i5-3470
3.20GHz CPU, 4GB RAM and 500GB hard disk with Microsoft Windows 7 operation system. Sample standard
AC

gray images of size 256 × 256 and 512 × 512 are utilized as plain images to carry out the experimental analyses.

4.1 Statistical attack analysis

Correlation coefficient, histogram and entropy depicted in the following subsections are the three most important
evaluation criteria for statistical attack analysis.

10
 ACCEPTED MANUSCRIPT

I 3 3 7 3 6

22 1 10

2 15 34

x1 , y1 , x 2 , y 2
P9
7 22 2 3 1 15 6 10 34

T
IP
Ry
7 1 7 7 8 1 8 2 5

CR
164 62 171 168 86 5 91 5 136

US DNAR
215 20 97 241 31 135 116 208 54
AN

115 42 202 89 73 130 47 213 190

M

Rx
4 6 8 1 6 7 4 1 3
ED

Loc
132 17 164 95 123 71 55 72 130 7 3 6 8 1 4 9 5 2
PT
CE

224 117 192 59 31 35 83 44 174

AC

83 44 174
Ic
192 224 31 83 192 35 44 224 59 174 31 117

35 59 117
Fig. 5 A simple example of the proposed encryption.

11
 ACCEPTED MANUSCRIPT

T
(a) The plain image (b) The cipher image (c) The decipher image

IP
Fig. 6 The plain, cipher and decipher image of the proposed scheme.

CR
4.1.1 Correlation coefficient test

The correlations between two horizontal, vertical and diagonal adjacent pixels are all called correlation coefficient
which is one of the most important evaluation method in the field of image encryption. The correlation coefficient

US
of cipher image encrypted by robust encryption scheme should be small and its ideal value is 0. The experiment
randomly selects 4000 pairs adjacent pixels in the plain image and cipher image respectively to calculate the
AN
correlation coefficient by
N · cov(x, y)
r xy = q , (5)
PN 2
PN 2
i=1 (xi − E x ) · i=1 (yi − E y )
M

1 PN
where E x = xi , cov(x, y) = E((x−E x )(y−Ey )) and x, y are the two of horizontal, vertical or diagonal adjacent
N i=1
pixels sequences. Table 3 shows the correlation coefficient of plain images and their cipher images encrypted by
ED

different encryption schemes. It can be seen that the proposed scheme has correlation coefficient of approximately
zero. Fig. 7 further demonstrates that the proposed algorithm transforms the linear correlation of a plain image into
an uncorrelated cipher image.
PT

4.1.2 Histogram analysis

CE

Histogram is another statistical feature in image encryption area. Image histogram shows the frequency of dif-
ferent pixel values of the image at different intensity values. For a robust encryption scheme, the histogram of
the cipher image demonstrates a balanced frequency in all of the pixel values. To analyze the histogram of cipher
AC

image encrypted by the proposed scheme, Fig. 8a and 8a are selected to be plain images and their histograms are
shown in Fig. 8b and 8e while their encrypted images are shown in Fig. 8c and 8f.
A quantity analysis of image histogram is represented in [35]. Its form is as follows:

n n
1 XX
Var(V) = (vi − v j )2 , (6)
2n2 i=1 j=1

where V is a one-dimensional array, and its ith element represents the frequency of gray value i in the testing image.

12
 ACCEPTED MANUSCRIPT

Table 3. Comparison of correlation coefficient in three directions for different encryption scheme

Image Size Direction

Plain image (256 × 256) horizontal 0.9579 0.9611 0.9584 0.9556 0.9592 0.9640
vertical 0.9503 0.9689 0.9510 0.4986 0.9682 0.9711
diagonal 0.9317 0.9391 0.9408 0.9291 0.9374 0.9485
(512 × 512) horizontal 0.9629 0.9579 0.9505 0.9591 0.9629 0.9671
vertical 0.9652 0.9615 0.9758 0.9531 0.9672 0.9744
diagonal 0.9296 0.9274 0.9283 0.9374 0.9407 0.9381

T
Ref. [35] (256 × 256) horizontal 0.0037 0.0062 0.0198 0.0023 0.0073 0.0059
vertical 0.0258 0.0074 0.0132 0.0019 0.0109 0.0041

IP
diagonal 0.0079 0.0009 0.0032 0.0011 0.0016 0.0028
(512 × 512) horizontal 0.0053 0.0037 0.0062 0.0008 0.0071 0.0016
vertical 0.0138 0.0014 0.0110 0.0021 0.0095 0.0048

CR
diagonal 0.0019 0.0008 0.0022 0.0005 0.014 0.0024
Ref. [36] (256 × 256) horizontal 0.0049 0.0054 0.0053 0.0061 0.0085 0.0060
vertical 0.0031 0.0089 0.0126 0.0116 0.0092 0.0058

(512 × 512)
diagonal
horizontal
vertical
diagonal
US
0.0079
0.0046
0.0025
0.0034
0.0021
0.0024
0.0104
0.0015
0.0005
0.0076
0.0078
0.0044
0.0018
0.0018
0.0112
0.0054
0.0024
0.0018
0.0083
0.0014
0.0016
0.0027
0.0021
0.0042
AN
Ref. [37] (256 × 256) horizontal 0.0021 0.0075 0.0145 0.0041 0.0073 0.0055
vertical 0.0218 0.0084 0.0084 0.0021 0.0216 0.0015
diagonal 0.0096 0.0011 0.0026 0.0009 0.0035 0.0041
(512 × 512) horizontal 0.0021 0.0033 0.0042 0.0018 0.0078 0.0065
M

vertical 0.0143 0.0067 0.0124 0.0026 0.0095 0.0047

diagonal 0.0029 0.0015 0.0031 0.0022 0.024 0.0017
Proposed (256 × 256) horizontal 0.0016 0.0028 0.0024 0.0056 0.0001 0.0026
ED

vertical 0.0059 0.0041 0.0013 0.0037 0.0031 0.0009

diagonal 0.0034 0.0010 0.0098 0.0032 0.0015 0.0052
(512 × 512) horizontal 0.0006 0.0025 0.0014 0.0032 0.0003 0.0029
vertical 0.0038 0.0050 0.0002 0.0016 0.0034 0.0033
PT

diagonal 0.0010 0.0012 0.0035 0.0023 0.0011 0.0062

CE

Var(V) is the variance of array V. The standard deviation shows the fluctuation of the measured value to the mean.
The larger the standard deviation, the greater the fluctuation of the measured value. Table 4 compares the variance
of different plain images and the corresponding cipher images encrypted by different schemes. The variance values
AC

in Ref. [35] are ranged in [5097, 5403] while the variance values of the proposed scheme are ranged in [940, 1103].
This illustrates that the average fluctuation in quantity of each gray value is around 72 pixels in Ref. [35] and around
31.7 pixels in the proposed scheme. Thus, the histogram of cipher image encrypted by the proposed scheme has a
more uniform distribution and can better resist the statistical attack.

13
 ACCEPTED MANUSCRIPT

T
IP
CR
(a) The plain image (b) The cipher image of (a) for improved scheme

US
AN
M
ED
PT

(c) The correlation of plain image (d) The correlation of cipher image
Fig. 7 Correlation plot of plain image and its cipher image

Table 4. Variance comparison of the plain and cipher 512 × 512 images encrypted by different schemes.
CE

Variance
AC

Plain image 548714.851 2821212.851 1673908.953 638716.843 1602343.507 745278.250

Ref. [35] 5403.283 5264.710 5381.261 5118.094 5097.985 5119.29
Ref. [36] 1033.789 889.634 953.203 1068.109 985.898 963.632
Ref. [37] 1105.716 1141.375 976.015 1043.214 973.765 1061.038
Proposed 946.67 1103.156 941.062 1027.593 926.914 1058.132

14
 ACCEPTED MANUSCRIPT

T
(a) The plain image (b) The cipher image (c) The decipher image

IP
CR
US
AN
(d) The plain image (e) The cipher image (f) The decipher image
Fig. 8 The histogram analysis of the proposed scheme.
M

4.1.3 Entropy analysis

Entropy is a significant measure to test the randomness of an image. The value of entropy for a real random
ED

image of 256 gray levels is 8. Entropy is calculated by

N −1
2X
1
H(G) = P(Gi ) log , (7)
PT

i=0
P(Gi )

where Gi is the gray value i, P(Gi ) is the probability of Gi in an image and N is the number of image pixels. Table
CE

5 shows the entropy of different plain images and the corresponding cipher images encrypted by different schemes
respectively. The maximum entropy in different sizes is displayed in bold. It can be seen from the table that most
of the entropy of cipher images encrypted by the proposed scheme is closer to 8 than that of the others. Therefore,
AC

the proposed scheme can generate a good random cipher image.

15
 ACCEPTED MANUSCRIPT

Table 5. Information entropy comparisons of plain and cipher images encrypted by different schemes.

Image Size
Plain image (256 × 256) 3.8212 3.2677 3.5024 3.6347 3.2921 3.5501
(512 × 512) 3.3097 3.1958 3.2639 3.1435 3.2827 3.2925
Ref. [35] (256 × 256) 7.9958 7.9974 7.9939 7.9975 7.9941 7.9938
(512 × 512) 7.9983 7.9991 7.9972 7.9994 7.9988 7.9981
Ref. [36] (256 × 256) 7.9972 7.9970 7.9971 7.9974 7.9973 7.9970
(512 × 512) 7.9990 7.9991 7.9993 7.9995 7.9992 7.9990

T
Ref. [37] (256 × 256) 7.9909 7.9912 7.9915 7.9913 7.9907 7.9912
(512 × 512) 7.9921 7.9925 7.9923 7.9924 7.9924 7.9922

IP
Proposed (256 × 256) 7.9974 7.9970 7.9975 7.9976 7.9971 7.9971
(512 × 512) 7.9993 7.9992 7.9993 7.9994 7.9994 7.9992

CR
4.1.4 Statistical attack for images with heterogeneous contents

US
Due to the prominent features in block data, complex texture, high redundancy and extreme correlation between
pixels, four images: the red blood cell image, the fingerprint image, the square image and the full black image
are selected to test the ability of anti-statistical attack of the proposed scheme. Table 6 lists correlation coefficient,
AN
variance and entropy of the four images, and we know from the table, no matter what the plaintext is, the correlation
coefficient of the ciphertext is always approximate to 0 and the entropy of the ciphertext is always approximate to
8. The variances of the four cipher images are ranged in [260,350], which illustrates that the average fluctuation
M

in quantity of each gray value is around 18.7 pixels in the proposed scheme. Therefore, the proposed scheme can
resist statistical attacks well.
ED

Table 6. Correlation coefficient, variance and entropy of the proposed scheme for four heterogeneous images.

Test Image Direction

PT

Correlation Plain image horizontal 0.9702 0.5442 0.9689 1.0000

coefficient vertical 0.9532 0.5759 0.9600 1.0000
diagonal 0.9221 0.4930 0.9326 1.0000
CE

Cipher image horizontal 0.0075 0.0023 0.0376 0.0118

vertical 0.0072 0.0284 0.0034 0.0059
diagonal 0.0155 0.0320 0.0154 0.0020
AC

Variance Plain image - 20308.42 70161.54 8357283 16711680

Cipher image - 347.93 280.64 293.21 264.50
Entropy Plain image - 6.7892 7.1423 1.6164 0.0000
Cipher image - 7.9968 7.9969 7.9967 7.9970

4.2 Differential attack

The performance of resistance to differential attack is usually tested by the NPCR (the number of pixel change
rate) and UACI (the unified average changing intensity) scores [38]. The two scores present the sensitivity of

16
 ACCEPTED MANUSCRIPT

cipher image to a tiny change of corresponding plain image. The larger the score, the more sensitive the encryption
scheme to plain image. The ideal value of NPCR is 1 and UACI is 0.334 [4]. The criteria of NPCR and UACI can
be calculated by
P
i, j D(i, j)
NPCR = × 100%,
P m×n 0
i, j |C − C |
UACI = × 100%,
255 × m × n (8)



 0
1, if C(i, j) , C (i, j),


T
Di, j = 



0, otherwise,

IP
where C and C 0 are two cipher images whose corresponding plain images are only one pixel different. In our test,

CR
6 different plain images are applied to compare the NPCR and UACI scores of different schemes, and 4 special
images are added to obtain the NPCR and UACI scores of the proposed scheme. The test results are shown in
Table 7 and Table 8 respectively. These tables present that the NPCR and UACI scores of the proposed scheme

US
are all approximate to their ideal values, and is better than the other schemes. Therefore, the proposed scheme can
resist differential attack and is more stronger than the other three.
AN
Table 7. NPCR and UACI of different images encrypted by different schemes.

Image Size
M

Ref. [35] (256 × 256) NPCR 0.984975 0.994176 0.990039 0.995193 0.992500 0.991051
UACI 0.329483 0.335254 0.331026 0.335851 0.333928 0.332517
(512 × 512) NPCR 0.993017 0.994883 0.992052 0.996304 0.992918 0.992394
ED

UACI 0.330026 0.333562 0.334390 0.335989 0.324162 0.333144

Ref. [36] (256 × 256) NPCR 0.996601 0.996532 0.996300 0.996217 0.996154 0.996081
UACI 0.334415 0.335098 0.334716 0.334159 0.335057 0.334125
(512 × 512) NPCR 0.996202 0.994350 0.996405 0.996152 0.996057 0.995966
PT

UACI 0.335323 0.335317 0.334109 0.335024 0.334726 0.335016

Ref. [37] (256 × 256) NPCR All the images’ NPCI are 5.0862 × 10−6
UACI All the images’ UACI are 1.9946 × 10−8
CE

Proposed (256 × 256) NPCR 0.996078 0.996231 0.996139 0.996200 0.996170 0.995925
UACI 0.334953 0.336358 0.335316 0.334169 0.336609 0.333822
(512 × 512) NPCR 0.996112 0.996261 0.996082 0.996002 0.996124 0.995903
UACI 0.335265 0.335782 0.335574 0.335079 0.335891 0.335281
AC

Table 8. NPCR and UACI of the proposed schemes.

Image
Test
NPCR 0.996282 0.996124 0.996460 0.995636
UACI 0.333934 0.334582 0.337275 0.333492

17
 ACCEPTED MANUSCRIPT

4.3 Exhaustive attack

Key sensitivity and key space are usually tested to analyze the ability of an encryption scheme in resisting the
exhaustive attack.

4.3.1 Key sensitivity

Key sensitivity analyzes the sensitivity of cipher image to a tiny alteration in secret key. The 2D-HSM is ex-
tremely sensitive to its parameters and initial values. To test the key sensitivity of the proposed scheme, the key

T
a1 (b1 , x10 , y01 ) is changed from 1.4 to 1.40000000000001 and maintaining the other keys to decrypt the cipher im-

IP
ages by the proposed scheme. Two plain images shown in Fig. 9 (a) and (e) are selected as test images, and their
decrypted images with the changed key a1 are shown in Fig. 9 (b) and (f), respectively. Similarly, a2 (b2 , x20 , y02 ) is

CR
changed from 2.7 to 2.70000000000001 and their decrypted images are shown in Fig. 9 (c) and (g), respectively.
From these figures, we know that when there is a tiny change to any secret key, the decryption will completely fail,
so the proposed scheme is extremely sensitive to secret keys.

US
AN
M

(a) The plain image (b) The decipher image with (c) The decipher image with (d) The decipher image with correct
changed key a1 changed key a2 keys
ED
PT
CE

(e) The plain image (f) The decipher image with (g) The decipher image with (h) The decipher image with correct
AC

changed key a1 changed key a2 keys

Fig. 9 Key sensitivity analysis

4.3.2 Key space

The secret key space should be large enough to resist exhaustive attack. In the proposed scheme, the secret keys
are the 2D-HSM parameters and initial values a1 , b1 , x10 , y01 , a2 , b2 , x20 , y02 , if each of them has a precision of 1014 ,
the key space is (1014 )8 = 10112 > 2100 , which is large enough to resist exhaustive attack [4].

18
 ACCEPTED MANUSCRIPT

4.4 Robustness against noise

When communicated in a real world, the encrypted image may be affected by various noises, such as Gauss noise,
salt and pepper noise, and even cropped by malicious. The robustness against noise for encrypted image is an vital
requirement for a cryptosystem. Six plain images: ‘peppers.jpg’, ‘blood.jpg’, ‘fingerprint.jpg’, ‘rectangle.jpg’ and
‘black.jpg’ with size 256 × 256 are selected to test the anti-noise performance of the proposed encryption scheme.
The six plain images are encrypted by the proposed scheme first, and then their cipher images are cropped by
25%, distorted by Gaussian noise with zero mean and 0.05 variance and affected by 5% salt and pepper noise,

T
respectively. Finally, these noisy cipher images are decrypted and the decipher images are shown in Table 9. From
the table, we can see that all the decipher images decrypted by their corresponding damaged cipher image are still

IP
highly visible, which indicates that the proposed scheme has good performance in robustness against clipping and
noise attacks.

CR
Table 9. Robustness against clipping and noise attacks.

Plain

Noise
type
image
US
AN
occlusion
M
ED

Gaussian
PT

salt & pepper

CE
AC

4.5 Computational complexity of the proposed scheme

The size of the plain image is denoted as m × n. The time consuming of the proposed scheme is mainly divided
into three parts. The first part is the key streams generation. This part generates four key streams x1 , y1 , x2 , y2 whose
lengths are all mn, and the complexity of the generation algorithm is O(m × n). The second is the diffusion part
which contains two DNA encoding steps and one step of XOR operation, and its complexity is also O(m × n). The
third part is the permutation which is the simplest part and has the same complexity of O(m × n). From the above

19
 ACCEPTED MANUSCRIPT

analyses, the computational complexity of the proposed scheme is still O(m × n) and linear, so the time consuming
of proposed scheme hinges on the size of plain image.

4.6 Further explanation of experimental phenomena

Experimental phenomena in “Differential attack” of Section 4.2 and “Key sensitivity” of Section 4.3.1 show that,
for our proposed scheme, with a very small change in the plain image or secret key, the content of the cipher image
can considerably change. However, in “Robustness against noise” of Section 4.4, the robustness against noise of the
cipher image basically means that with high distortions of the cipher image, the plain image can still be recovered

T
with a very good version, which seems to contradict the differential analysis. Hence, much more explanations of our

IP
results in good key sensitivity, resisting differential attack and robustness against noise are provided for someone
who believe the three performances seem to contradict each other and cannot exist in an encryption scheme at the

CR
same time. We refute the above opinion by presenting the preconditions of the three performances.

• Key sensitivity is proposed to resist exhaustive attack. An attacker intercepts a cipher image and tries to

US
exhaust key to obtain a visible image. The attacker who decrypts cipher image with different secret keys
may obtain almost the same decipher image for an encryption scheme which is not sensitive to keys, then
the probability of breaking the scheme by exhaustive attack will be greatly reduced and thus the encryption
AN
scheme is not good.

• Differential attack is a kind of chosen plaintext attack which defines an attacker who has the ability to access
M

the encryption machine and tries to crack a cipher image without secret keys. For this attack, the attacker
adaptively chooses some plain images, constructs the corresponding cipher images through accessing the
encryption machine and then compares the acquired cipher images to extract some unknown terms about the
ED

encrypted system. Note that the attacker can only choose the plain images that are different from the plain
image of the cipher he wants to crack. Thus, differential attack is proposed to measure the impact of one
PT

pixel value changed in plain image on the cipher image, rather than the plain image on the decipher image.

Consider two extreme cases: one is that only one of the cipher image pixel changed when one pixel of the
CE

plain changed (NPCR ≈ 0), then the relationship between the plaintext and the ciphertext is linear and can be
quickly found out by the attacker. Therefore, the attacker can crack other ciphertext by finding relationship.
Another is that all of the cipher pixel changed when one pixel of the plain changed (NPCR = 1). This
AC

situation shows a complex nonlinear relation between the plaintext and the ciphertext, and thus the attacker
is confused to the effect of plaintext changed, which also represents the failure of differential attack. Thus,
the greater the value of NPCR, the more difficult it is for attackers to judge the relationship between the
plaintext and ciphertext and the more effective for the cryptosystem to resist differential attack. The ideal
value of NPCR is 1.

• Anti noise attacks. Noise attack is the influence of the natural environment or the intentional destruction of
the attacker during the transmission of the cipher image. The former is a natural phenomenon while the latter

20
 ACCEPTED MANUSCRIPT

is a man-made attack to prevent the receiver from obtaining correct image. The key point of the anti-noise
attack is how to effectively restore the plaintext.

We know that from the above, these three performances are generated in different applications and the compar-
isons between them are not on the same level. Thus, there exists no contradiction between the three performances.
That is to say, a good encryption scheme can possess good key sensitivity and strong robustness against differential
attack and noise attack at the same time.

T
5 Conclusions

IP
This paper has proposed a new image encryption scheme based on chaotic map and DNA approach. In order to
extend the scope of secret keys (chaotic parameters), a new 2D chaotic map named 2D-HSM has been designed.

CR
The map is generated by combining the Hénon map with the Sine map. Trajectory, the largest Lyapunov exponent
and bifurcation diagram are provided to evaluate that the new map has better ergodicity, wider chaotic range, and

US
its outputs are more unpredictable. A DNA approach is used to improve the computing efficiency of the encryption
scheme because it encodes each of a 8bit pixel value into four DNA elements which can be diffused by using DNA
XOR operation. Experimental simulation and comparison have also verified the security of the proposed scheme
AN
from four aspects: the exhaustive attack, the statistical attack, the differential attack and the noise attack. The
scheme has a large key space and is extremely sensitive to its keys, thus, it can resist exhaustive attack. Histogram
of the scheme is uniform, correlation coefficient is close to 0 and the entropy value is close to 8 for 8bit image, thus,
M

the scheme can resist statistical attack. Both UACI and NPCR values approach their ideal value, and the missing
cipher image can still be recovered well, which illustrates that the proposed scheme can resist differential attack
ED

and noise attack well.

All of the above can prove that the proposed scheme is efficient and practical in communications, but there are
still spaces to be explored and improved, one is that the security of the proposed scheme is only characterized by
PT

a variety of indicators, but its security proof is difficult to realize; another is that the input and output of the image
encryption algorithm in this paper have the same amount of data; the other is that the proposed scheme is designed
CE

mainly for gray image, and color image and multimedia data must first be converted to the same pattern of gray
images, and then encrypted with the scheme. In the future, we intend to study the security proof of the image
encryption algorithm, and consider the image encryption with compression and convert the encryption scheme into
AC

multimedia field.

Acknowledgements

This work is supported by National Key Research and Development Program of China (Grant no.
2016YFB0800601), Natural Science Foundation of China (Grant no. 61472331, 61772434).

21
 ACCEPTED MANUSCRIPT

References

[1] FIPS PUB 46, Data encryption standard (DES), 1999.

[2] FIPS PUB 197, Advanced encryption standard (AES), 2001.

[3] Fridrich J. Image encryption based on chaotic maps[J]. 1997, 2(2):1105-1110.

[4] Wu J, Liao X, Yang B. Color image encryption based on chaotic systems and elliptic curve Elgamal scheme[J].
Signal Processing, 2017, 141:109-124.

T
[5] Pak C, Huang L. A new color image encryption using combination of the 1D chaotic map[J]. Signal Process-

IP
ing, 2017, 138:129-137.

CR
[6] Patidar V, Pareek N K, Sud K K. A new substitution-diffusion based image cipher using chaotic standard and
logistic maps[J]. Communications in Nonlinear Science & Numerical Simulation, 2009, 14(7):3056-3075.

US
[7] Wang X, Guo K. A new image alternate encryption algorithm based on chaotic map[J]. Nonlinear Dynamics,
2014, 76(4):1943-1950.
AN
[8] Gao T, Chen Z. A new image encryption algorithm based on hyper-chaos[J]. Physics Letters A, 2008,
372(4):394-400.

[9] Zhou G, Zhang D, Liu Y, et al. A novel image encryption algorithm based on chaos and line map[J]. Neuro-
M

computing, 2015, 169:150-157.

[10] Zhu H, Zhang X, Yu H, et al. An image encryption algorithm based on compound homogeneous hyper-chaotic
ED

system[J]. Nonlinear Dynamics, 2017:1-19.

[11] Peng Z P, Wang C H, Lin Y, et al. A novel four-dimensional multi-wing hyper-chaotic attractor and its appli-
PT

cation in image encryption[J]. Acta Physica Sinica, 2014, 63(24):240506-240506.

[12] Enayatifar R, Abdullah A H, Isnin I F. Chaos-based image encryption using a hybrid genetic algorithm and a
CE

DNA sequence[J]. Optics & Lasers in Engineering, 2014, 56(5):83C93.

[13] Chen J, Zhu Z L, Zhang L B, et al. Exploiting self-adaptive permutation-diffusion and DNA random encoding
AC

for secure and efficient image encryption[J]. Signal Processing, 2018, 142:340-353.

[14] Zhang Y, Xiao D. Double optical image encryption using discrete Chirikov standard map and chaos-based
fractional random transform[J]. Optics & Lasers in Engineering, 2013, 51(4):472-480.

[15] Zhang Y Q, Wang X Y. A new image encryption algorithm based on non-adjacent coupled map lattices[J].
Applied Soft Computing, 2015, 26:10-20.

22
 ACCEPTED MANUSCRIPT

[16] Zhang H, Wang X Y, Wang S W, et al. Application of coupled map lattice with parameter q, in image encryp-
tion[J]. Optics & Lasers in Engineering, 2017, 88:65-74.

[17] Wang Y, Zhao Y, Zhou Q, et al. Image encryption using partitioned cellular automata[J]. Neurocomputing,
2017. DOI: 10.1016/j.neucom.2017.09.068

[18] Hanis S, Amutha R. Double image compression and encryption scheme using logistic mapped convolution
and cellular automata[J]. Multimedia Tools & Applications, 2017(2):1-16.

T
[19] Zhou N, Yang J, Tan C, et al. Double-image encryption scheme combining DWT-based compressive sensing
with discrete fractional random transform[J]. Optics Communications, 2015, 354:112-121.

IP
[20] Zhang D, Liao X, Yang B, et al. A fast and efficient approach to color-image encryption based on compressive

CR
sensing and fractional Fourier transform[J]. Multimedia Tools & Applications, 2017:1-18.

[21] Gehani A, LaBean T H, Reif J H. DNA-based cryptography[J]. Dimacs Series In Discrete Mathematics and
Theoretical Computer Science, 2000,54:233-249
US
[22] Zhang Q, Guo L, Wei X. Image encryption using DNA addition combining with chaotic maps[J]. Mathemat-
AN
ical & Computer Modelling, 2010, 52(11C12):2028-2035.

[23] Su X, Li W, Hu H. Cryptanalysis of a chaos-based image encryption scheme combining DNA coding and
entropy[J]. Multimedia Tools & Applications, 2016:1-13. DOI: 10.1007/s11042-016-3800-9
M

[24] Chen L, Ma B, Zhao X, et al. Differential cryptanalysis of a novel image encryption algorithm based on chaos
and Line map[J]. Nonlinear Dynamics, 2017, 87(3):1797-1807.
ED

[25] Yap W S, Phan C W, Yau W C, et al. Cryptanalysis of a new image alternate encryption algorithm based on
chaotic map[J]. Nonlinear Dynamics, 2015, 80(3):1483-1491.
PT

[26] Jeng F G, Huang W L, Chen T H. Cryptanalysis and improvement of two hyper-chaos-based image encryption
schemes[J]. Signal Processing Image Communication, 2015, 34:45-51.
CE

[27] Chengqing Li, Shujun Li, Kwok-Tung Lo. Breaking a modified substitution-diffusion image cipher based on
chaotic standard and logistic maps[J]. Communications in Nonlinear Science & Numerical Simulation, 2011,
AC

16:837C843.

[28] Wu J, Liao X, Yang B. Cryptanalysis and enhancements of image encryption based on three-dimensional bit
matrix permutation[J]. Signal Processing, 2017, 142:292-300.

[29] Short K M. Steps toward unmasking secure communications[J]. International Journal of Bifurcation and
Chaos, 1994, 4(04): 959-977.

23
 ACCEPTED MANUSCRIPT

[30] Wu Y, Yang G, Jin H, et al. Image encryption using the two-dimensional logistic chaotic map[J]. Journal of
Electronic Imaging, 2012, 21(1): 013014-1-013014-15.

[31] Hua Z, Zhou Y, Pun C M, et al. 2D Sine Logistic modulation map for image encryption[J]. Information
Sciences, 2015, 297: 80-94.

[32] Hua Z, Zhou Y. Image encryption using 2D Logistic-adjusted-Sine map[J]. Information Sciences, 2016, 339:
237-253.

T
[33] Shevchenko I I. Lyapunov exponents in resonance multiplets[J]. Physics Letters A, 2014, 378(1): 34-42.

IP
[34] Liao X, Xiao D. Theory and applications of chaotic cryptography[M]. Science Press, 2009.
http://www.abook.cn/pdf/4302.0101.pdf

CR
[35] Enayatifar R, Abdullah A H, Isnin I F, et al. Image encryption using a synchronous permutation-diffusion
technique[J]. Optics & Lasers in Engineering, 2017, 90:146-154.

US
[36] Niyat A Y, Moattar M H, Torshiz M N. Color image encryption based on hybrid hyper-chaotic system and
cellular automata[J]. Optics & Lasers in Engineering, 2017, 90:225-237.
AN
[37] Li C, Luo G, Qin K, et al. An image encryption scheme based on chaotic tent map[J]. Nonlinear Dynamics,
2017, 87(1):127-133.
M

[38] Wu Y. NPCR and UACI randomness tests for image encryption[J]. Cyber Journals: Multidisciplinary Journals
in Science and Technology, Journal of Selected Areas in Telecommunications (JSAT, 2011.)
ED
PT
CE
AC

24