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.2020.2988880, IEEE Access

Image encryption based on improved Lorenz

system
Chengye Zou1, Qiang Zhang2,3, Xiaopeng Wei3 and Chanjuan Liu2
1
School of Mathematics and Statistics, Anyang Normal University, Anyang 455000, China
2
Faculty of Electronic Information and Electrical Engineering, Dalian University of Technology, Dalian 116024, China
3
Key Laboratory of Advanced and Intelligent Computing, Dalian University, Ministry of Education, Dalian 116622, China
Corresponding author: Qiang Zhang (e-mail: [email protected]).
This work was supported in part by the National Natural Science Foundation of China under Grant Nos: 61425002, 61751203, and 61772100, Program for
Changjiang Scholars and Innovative Research Team in University under Grant No. IRT_15R07, Program for Liaoning Innovative Research Team in
University under Grant No. LT2015002.

ABSTRACT This paper proposes a new method to strengthen the nonlinear kinetic complexity and
ergodicity of Lorenz system. Through the analysis of auto-correlation, frequency distribution, approximate
entropy and information entropy, the improved Lorenz system has better dynamical properties than Lorenz
system. According to NIST-800-22 test results, the chaotic sequences generated from proposed system have
passed all random tests, which denotes that the improved Lorenz system is applicable to chaotic encryption.
Once the plaintext image is color, the size of scramble image is three times as big as plaintext image, so that
a lot more pixel information participant in permutation and diffusion to get better encryption results.
Simulation results show that the image encryption scheme provides good security and high capacity to resist
common attacks.

INDEX TERMS ergodicity, image encryption, NIST-800-22 test, Lorenz system, improved Lorenz system

I. INTRODUCTION have designed a cryptographically strong systems based on

Since Lorenz proposed the first two scrolls system in 1963 proposed S-boxes and Lorenz chaotic system [24]. Girdhar
[1], chaotic systems have attracted attention of scientific and Kumar hybridized Lorenz system and Rossler system to
community due to their vast applications in many areas such generate the random sequence, and applied the rules of DNA
as physics, biology, complex networks, economics and so on cryptosystem to encrypt image [25].
[2-6]. Non-predictability, ergodicity, random-like behavior, However, dimension, complexity and ergodicity of
and high sensitivity to initial conditions are the main Lorenz system are limited, therefore the original Lorenz
characteristics of chaotic systems, which can be used in system is improved or modified in many investigations to
encryption and secure communication [7-20]. Generally, achieve better encryption results. Kaur and Kumar generated
diffusion and permutation are two significant steps in chaotic six random sequences by Lorenz-like chaotic system with
image encryption. In permutation phase, strong correlations varying bifurcation parameter [26]. It is obvious that, hyper-
between adjacent pixels are masked. In the diffusion phase, chaotic systems possess more complex dynamical
pixel values are replaced by mutual diffusion among characteristics and higher randomness than low-dimensional
different pixels, so that the important and valid information chaotic systems, so that hyper-Lorenz system is more
of plaintext-image are concealed. In many proposed chaotic applicable to image encryption [27-29]. Usman et al.
image encryption algorithms, Lorenz system as pseudo proposed Walsh compressed quantum spinning chaotic
random number generator (PRNG) to generate chaotic Lorenz system to satisfy the need of fast computing and
sequences. Obaida et al. applied Lorenz system to digital quantum encryption [27]. Ran et al. have coupled two
image encryption, and embedded hash value of the plain Lorenz system and obtained hyper-Lorenz system, which is
image in the proposed cryptosystem to increase the security injected impulse during the iteration process in order to
level [21]. Amir has proposed a robust image encryption counter the degeneration of dynamics [30].
scheme for low profile applications based on chaotic Lorenz In this paper, the motivations of our work are included as
system [22]. Younas combined inverse left almost semi follows:
group with Lorenz system in image encryption scheme to (1) In order to strengthen the complexity and ergodicity of
obtain better confusion and diffusion effect through a Lorenz system, we have added delay coupling and mod
modern substitution-permutation network [23]. Hussain et al. function to Lorenz system, and enhanced complexity and

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

ergodicity of Lorenz system effectively according to the where 𝑥̂(𝑡), 𝑦̂(𝑡) and 𝑧̂ (𝑡) are output sequences.
analysis of auto correlation, frequency distribution,
approximate entropy and information entropy
(2) To avoid attacker gain the random value of the
proposed system through the selection of different plaintext
images, the keys derive from plaintext image. Furthermore,
the key space can be enlarged greatly by increase of key
number.
(3) To make the information of plaintext image participant
the permutation and diffusion more sufficiently, the
scramble image has three times the size of plaintext image
for color plaintext image. This permutation operation will
not increase the burden of transmission or storage, because
we just transfer the information of R, B,G components of
plaintext image to the scramble image. If the the plaintext
image is gray, scramble image has the same size as plaintext
image. The experimental results showed that our proposed FIGURE 1. Mechanism of our proposed system.
encryption scheme has high security.
The rest of the paper is organized as follows: Section II B. Experiment Results
presents basic mechanism and analysis of improved Lorenz
system. Our encryption and decryption algorithm are
depicted in Section III. Simulation results and security
analyses are provided in Section IV and V respectively.
Section VI and VII present the statistical and robust analysis
respectively. Finally, this paper is concluded in Section VIII.

II. Improved Lorenz system

A. Basic mechanism
Lorenz system is one of the most widely used chaotic
system, especially in image encryption. The mathematical
model of Lorenz system can be described as follow:
𝑑𝑥(𝑡)
= 𝑎(𝑦(𝑡) − 𝑥(𝑡)) (a)
𝑑𝑡
𝑑𝑦(𝑡)
= 𝑏𝑥(𝑡) − 𝑥(𝑡)𝑧(𝑡) − 𝑦(𝑡) (1)
𝑑𝑡
𝑑𝑧(𝑡)
{ 𝑑𝑡 = 𝑥(𝑡)𝑦(𝑡) − 𝑐𝑧(𝑡)
where chaos of Eq. (1) is determined by the parameters 𝑎,
𝑏 and 𝑐. As shown in Figure 1, we modified the form of Eq.
(1) as:
d𝑥(𝑡)
= 𝑎(𝑦̂(𝑡) − 𝑥̂(𝑡))
d𝑡
d𝑦(𝑡)
= 𝑏𝑥̂(𝑡) − 𝑥̂(𝑡)𝑧̂ (𝑡) − 𝑦̂(𝑡) (2)
d𝑡
d𝑧(𝑡)
{ d𝑡 = 𝑥̂(𝑡)𝑦̂(𝑡) − 𝑐𝑧̂ (𝑡)
The form of 𝑥̂(𝑡), 𝑦̂(𝑡) and 𝑧̂ (𝑡) is designed as
𝑥̂(𝑡) = (mod(𝑓(𝑥(𝑡), 𝑥(𝑡 − 𝜏), t)), 1)
(b)
{𝑦̂(𝑡) = (mod(𝑓(𝑦(𝑡), 𝑦(𝑡 − 𝜏), t)), 1) (3)
FIGURE 2. Phase diagram of Lorenz system and proposed system with
𝑧̂ (𝑡) = (mod(𝑓(𝑧(𝑡), 𝑧(𝑡 − 𝜏), t)), 1) the same initial value, (a) Lorenz system, (b) proposed system.
where the function 𝑓(𝑥(𝑡), 𝑥(𝑡 − 𝜏), t), 𝑓(𝑦(𝑡), 𝑦(𝑡 − 𝜏), t)
and 𝑓(𝑧(𝑡), 𝑧(𝑡 − 𝜏), t) are described as Eq. (4), and 𝜏 > 0 In experiment, the initial values of Lorenz system are
is delayed time. fixed as 𝑥0 = 0.01 , 𝑦0 = 0.02 and 𝑧0 = 0.02. The initial
𝑥(𝑡)+𝑥(𝑡−𝜏) values of improved system are the same as Lorenz system.
𝑓(𝑥(𝑡), 𝑥(𝑡 − 𝜏), t) = + sin(𝑡)
2 Taking parameters 𝑎 = 20, 𝑏 = 50 and 𝑐 = 8. Delayed time
𝑦(𝑡)+𝑦(𝑡−𝜏)
𝑓(𝑦(𝑡), 𝑦(𝑡 − 𝜏), t) = + sin(𝑡) (4) 𝜏 is fixed as 𝜏 = 1. As shown in Figure 2, our proposed
2
𝑧(𝑡)+𝑧(𝑡−𝜏)
system can break the two scrolls attractor structure totally
{ 𝑓(𝑧(𝑡), 𝑧(𝑡 − 𝜏), t) = + sin(𝑡) and make the attractor as the noise-like pattern, which
2
2

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

denotes that proposed system has better ergodicity than

Lorenz system. As shown in Figure 3, temporal evolution of
proposed system oscillates more seriously than Lorenz
system. The trajectories of improved system are random-like
and ergodic with no apparent cycle in the phase space.

FIGURE 4. Approximate entropy analysis of Lorenz system, improved

Lorenz system, Huang’s system and Ran’s system

As shown in Figure 4, the AE of proposed system is larger

than Lorenz system, Lorenz-like system [26] and coupled
Lorenz system [27] with different precisions, which
(a)
demonstrates that proposed system can strengthen the
complexity of Lorenz system greatly.

D. Information entropy
Shannon proposed information entropy in 1948 [32], and
the degree of confusion can be characterized by information
entropy. The system is more ordered, information entropy is
lower. It is defined as:
𝑁

𝛨(𝑠) = − ∑ 𝑃(𝑠𝑖 )log 2 (𝑃(𝑠𝑖 )) (7)

𝑖=1

(b)
FIGURE 3. Temporal evolution of system: (a) Lorenz system; (b)
proposed system

C. Approximate Entropy
The complexity of time series can be measured by
approximate entropy (AE) [31], given a time-series of data
{𝑢(𝑖), 𝑖 = 1, 2, ⋯ , 𝑁}, reconstruct this series as:
𝑋𝑖 = {𝑢(𝑖), 𝑢(𝑖 + 1), ⋯ , 𝑢(𝑖 + 𝑚 − 1)} , 𝑖 = 1, 2, ⋯ , 𝑛 ,
𝑛 = 𝑁 − 𝑚 + 1.
where 𝑋𝑖 is a 𝑚-dimentional vector. Calculate the distance
between 𝑥𝑖 and 𝑥𝑗 as
𝑑[𝑥(𝑖) − 𝑥(𝑗)] = max (|𝑢(𝑖 + 𝑘) − 𝑢(𝑗 + 𝑘)|) (5) FIGURE 5. Analysis of information entropy of Lorenz system, improved
𝑘=0,1,⋯,𝑚−1 Lorenz system, Huang’s system and Ran’s system
Given a threshold 𝑟 ∈ [0.2, 0.3] , Let 𝛫 be the number
satisfying the equation 𝑑𝑖𝑗 ≤ 𝑟 × SD , where SD is the where 𝑠 is information source, 𝑁 demonstrates the number
standard value of the sequence, then we define of states, 𝑃(𝑠𝑖 ) indicates the probability of occurrence of 𝑠𝑖 .
𝑁−𝑚+1 The sequences 𝑥, 𝑦, 𝑧, 𝑥̂, 𝑦̂ and 𝑧̂ are hypothesized to 256
1
𝜙 𝑚 (𝑟) = ∑ ln 𝐶𝑖𝑚 (𝑟) (6) states, in theory the maximum information entropy of each
𝑁−𝑚+1 sequence is log 2 256 = 8. As shown in Figure 5, all entropy
𝑖=1
𝛫
where 𝐶𝑖𝑚 (𝑟) = , then the AE of a sequence can be value of our proposed system is close to 7.8, which is bigger
𝑁−𝑚
than the other systems.
calculated as 𝜙 − 𝜙 𝑚+1 .
𝑚

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

(a) (b) (c)

(d) (e) (f)

FIGURE 6. Auto-correlation functions of Lorenz system and proposed system: (a), (d) and (c) Lorenz system; (d), (e) and (f) proposed system.

(a) (b) (c)

(d) (e) (f)

FIGURE 7. Frequency distributions: (a), (d) and (c) Lorenz system; (d), (e) and (f) proposed system.

where 𝑥𝑖 and 𝑦𝑖 represent the 𝑖th sequences, 𝑥̅ and 𝑦̅ are the

E. Auto correlation average of sequence 𝑥 and 𝑦 respectively. Figure 6 (a), (b)
Auto correlation (AC) is also a significant factor used to and (c) show the auto-correlations of Lorenz system. As
evaluate the dynamical characteristics of a chaotic system. shown in Figure 6 (c), the auto-correlation of 𝑧 is much
The correlation is defined as follow [33]: stronger than 𝑥 and 𝑦. As shown in Figure 6 (d), (e) and (f),
∑𝑁𝑖=1(𝑥𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦
̅) the autocorrelation of adjacent orbits of the proposed system
𝜒𝑥,𝑦 = (8) is weaker than the Lorenz system. Auto-correlation function
𝑁 (𝑥
√∑𝑖=1 𝑖 − 𝑥̅ )2 (𝑦𝑖 − 𝑦̅)2
of improved Lorenz system has a small main peak, and the
4

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

side lobes are rarely low and even virtually invisible, which We apply sequencing index function to 𝑆 as follow:
demonstrates that our proposed system are good random. [𝐻, 𝑞] = sort(𝑆) (9)
where[∙,∙] = sort(∙) is the sequencing index function, 𝐻 =
F. Frequency distribution [ℎ1 , ℎ2 , ⋯ , ℎ𝑀×3𝑁 ] is the new sequence after ascending to 𝑆,
Figure 7 depicts the frequency distribution of Lorenz and 𝑞 = [𝑞1 , 𝑞2 , ⋯ , 𝑞𝑀×3𝑁 ] is the index value of 𝐻.
system and proposed system, where the whole interval is Then, we have
divided into 500 equal sub-intervals. It can be seen that the 𝑀×3𝑁

distribution of proposed system is much more homogeneous sum = ∑ ℎ𝑚 × 𝑞𝑚 (10)

than Lorenz system. Therefore, the proposed perturbation 𝑚=1
method can resist frequency attack effectively. We utilize Eq. (11) to generate 𝛫 keys as
𝑑𝑘 = mod(sum, 𝑘 + 0.1) (11)
G. NIST-800-22 statistical tests where 𝑘 ∈ [1,2, ⋯ , 𝛫], and average value 𝑑 = ∑𝐾 𝑘=1 𝑘 /𝛫
𝑑
The NIST-800-22 test was designed to assess the is obtained, 𝛫 is assumed as 𝛫 ≥ 3 to satisfy the inequality
performance of Pseudo Random Number Generator (PRBG) 1014𝛫 > 2100 .
[34]. We can judge whether or not a chaotic binary sequence In this article, the keys are associated with plaintext
is suitable to cryptographic algorithm based on the results of images, if plaintext images are different, the keys will be
NIST test. The NIST-800-22 test is made up of 15 test updated. The update of keys may bring encryption
methods, including frequency test, run test, approximate inconvenience, but which will improve security of algorithm,
entropy test, random excursions test, etc. Randomness of test because the attacker cannot obtain the random value of the
sequence can be measured by 𝑝 value. If 𝑝 ≥ 0.01 , the improved system by selecting different plaintext images [35].
sequence is random. If 𝑝 < 0.01, the sequence is not random. The initial values, and parameter 𝑎, 𝑏 and 𝑐 of the Eq. (2)
If 𝑝 = 1, the sequence is completely random. If 𝑝 = 0, the can be calculated by the follows:
sequence is not at all random. Moreover, 𝑝 value is bigger, 𝑥0 = 𝑑1 mod 1
the randomness of the sequence is better. Table 1 shows 𝑦0 = 𝑑2 mod 1
NIST tests of PRNG, which can be found sequence of 𝑧0 = 𝑑3 mod 1
improved Lorenz system passes all random tests. 𝑎 = 20 + 𝑑1 mod 0.1
TABLE 1
𝑏 = 50 + 𝑑2 mod 0.1
TEST RESULTS BY NIST TEST SUITE { 𝑐 = 8 + 𝑑3 mod 0.1
Test Item Means of 𝒑-Value Result If plaintext image is gray, 𝐻 = [ℎ1 , ℎ2 , ⋯ , ℎ𝑀×𝑁 ] , 𝑞 =
Approximate Entropy 0.48636 Passed [𝑞1 , 𝑞2 , ⋯ , 𝑞𝑀×𝑁 ], and sum is calculated as
Block Frequency 0.55117 Passed 𝑀×𝑁
Cumulative Sums 0.49226 Passed
FFt 0.51597 Passed
sum = ∑ ℎ𝑚 × 𝑞𝑚
Frequency 0.50321 Passed 𝑚=1
Linear Complexity 0.40560 Passed Step 2. Generate three sequences 𝑥̂, 𝑦̂ and 𝑧̂ with the length
Longest Run 0.51564 Passed of 𝑀 × 𝑁 and reshape 𝑥̂, 𝑦̂ and 𝑧̂ as Eq. (12), then spliced
Nonoverlapping Template 0.50127 Passed them together into two big matrixes 𝑊1 and 𝑊2 as Eq. (13):
Overlapping Template 0.48756 Passed
Random Excursions 0.52112 Passed 𝑋1 = reshape(𝑥̂ × 102 , [𝑀, 𝑁])
Random Excursions Variant 0.50540 Passed {𝑋2 = reshape(𝑦̂ × 102 , [𝑀, 𝑁]) (12)
Ranks 0.48194 Passed
Runs 0.56419 Passed
𝑋3 = reshape(𝑧̂ × 102 , [𝑀, 𝑁])
Serial Test Passed 𝑊 = [𝑋1 , 𝑋2 , 𝑋3 ]
0.52448 { 1 (13)
Maurer’s Universal 0.67773 Passed 𝑊2 = [𝑋2 , 𝑋3 , 𝑋1 ]
If plaintext image is gray, matrixes 𝑊1 and 𝑊2 are built as:
III. Encryption and decryption algorithms 𝑊 = [𝑋1 ]
{ 1
𝑊2 = [𝑋2 ]
A. Encryption algorithm Step 3. Create a new matrix by using the following method:
Suppose that a plaintext image 𝑃 with a size of 𝑀 × 𝑁 is 𝑊{𝑖, 𝑗} ← 𝑊1 {𝑖, 𝑗} × 𝑊2 {𝑖, 𝑗} (14)
given. If plaintext image is a color image, 𝑃 is converted into where 𝑖 = 1,2, ⋯ , 𝑀 and 𝑗 = 1,2, ⋯ ,3𝑁. If plaintext image
R, G and B components, and we can obtain three matrixes is gray, 𝑗 = 1,2, ⋯ , 𝑁.
𝑃r , 𝑃g and 𝑃b , the size of each color’s matrix is 𝑀 × 𝑁, Step 4. Replace the values of the first column of 𝑊 with 𝐷,
where the pixel values range from 0 to 255. The encryption and obtain a new matrix 𝛹, where 𝐷 is calculated as:
process is as follows. 𝐷 = (𝑑×1014 ) mod ξ (15)
Step 1. When plaintext image is color, construct a big matrix and
𝑆 = [𝑃r , 𝑃g , 𝑃b ], so that the information of 𝑃r , 𝑃g and 𝑃b is
∑𝐾 𝑘=1(𝑑𝑘 − 𝑑)
2
transferred to a big matrix 𝑆; when plaintext image is gray, 𝑑=√
the matrix 𝑆 is built as 𝑆 = 𝑃 to avoid additional burdens of 𝐾
transmission or storage. ξ=min{|𝑋1 |max , |𝑋2 |max , |𝑋3 |max }
5

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

Step 5. Apply sequencing index function to each row of 𝛹 then reshape three matrixes 𝑋̃1 , 𝑋̃2 and 𝑋̃3 as
according to Eq. (9), and obtain the index matrix 𝑈, then 𝑋̃1 = reshape(𝑋̃1 , [𝑀, 𝑁])
select the pixels of 𝑆 according to the follow law: {𝑋̃2 = reshape(𝑋̃2 , [𝑀, 𝑁]) (19)
𝑇1 {𝑖, 𝑗} ← 𝑆{𝑖, 𝑈𝑖,𝑗 } (16)
𝑋̃3 = reshape(𝑋̃3 , [𝑀, 𝑁])
Step 6. Replace the values of the first row of 𝑊 with 𝐷, and Step 9. We build two big matrixes 𝑊 ̃1 and 𝑊
̃2 according to
obtain a new matrix Φ.
Eq. (13) by 𝑋̃1 , 𝑋̃2 and 𝑋̃3 , and regenerate the matrix 𝑊̃ as:
Step 7. Apply sequencing index function to each column of ̃ ̃ ̃
{𝑖, {𝑖,
𝑊 𝑗} ← (𝑊1 𝑗} × 𝑊2 𝑗}) mod 𝜃 {𝑖, (20).
Φ according to Eq. (9), and obtain the index matrix 𝑉, then
select the pixels of 𝑇1 according to Eq. (17): The manipulation diffusion is defined as
𝐶i = (𝑇𝑖 ⨁𝑇𝑀×3𝑁 + 𝑊 ̃𝑖 )mod 𝜃 if 𝑖 = 1
𝑇{𝑖, 𝑗} ← 𝑇1 {𝑉𝑖,𝑗 , 𝑗} (17)
As shown in Figure 8, an example with the color image size {𝐶𝑖 = 𝑇𝑖 ⨁𝑇𝑖+1 ⨁𝐶𝑖−1 ⨁𝑊̃𝑖 if 1 < 𝑖 < 𝑀 × 3𝑁
of 3 × 4 is used to better explain the process of permutation. ̃
𝐶𝑖 = (𝑇𝑖 ⊕ 𝐶𝑖−1 + 𝑊𝑖 )mod 𝜃 if 𝑖 = 𝑀 × 3𝑁
A good diffusion property of encryption should make (21)
cipher-text image sensitive to the change of plaintext. If plaintext image is gray, the diffusion operation is defined
Usually, XOR operation is a manipulation on the level of bit, as
but addition operation is a manipulation on the level of pixel. ̃𝑖 )mod 𝜃 if 𝑖 = 1
𝐶i = (𝑇𝑖 ⨁𝑇𝑀×3𝑁 + 𝑊
We combine XOR operation with addition operation to
enhance the effectiveness of diffusion. { 𝐶𝑖 = 𝑇𝑖 ⨁𝑇𝑖+1 ⨁𝐶𝑖−1 ⨁𝑊 ̃𝑖 if 1 < 𝑖 < 𝑀 × 𝑁
Step 8. We generated three chaotic sequences as: 𝐶𝑖 = (𝑇𝑖 ⊕ 𝐶𝑖−1 + 𝑊̃𝑖 )mod 𝜃 if 𝑖 = 𝑀 × 𝑁
𝑥̂
𝑋̃1 = ⌊( × 1014 ) mod 𝜃⌋ where ⊕ is the bit wise XOR operation, and 𝜃 is the
𝑑
𝑦̂ maximum value of pixel in plaintext-image.
𝑋̃2 = ⌊( × 1014 ) mod 𝜃⌋ (18)
𝑑
𝑧̂
̃ 14
{𝑋3 = ⌊(𝑑 × 10 ) mod 𝜃⌋

Figure 8. An example of permutation through the color image of size 𝟑 × 𝟒.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

FIGURE 9. Flowchart of the encryption process.

(a) (b) (c) (d) (e)

FIGURE 10. Experimental results obtained using the proposed scheme: (a) Four different kinds of plaintext images; (b) histograms of (a); (c)
encrypted image of (a); (d) histograms of (c); (e) decrypted image of (c).

Step 10. Let 𝐶 = reshape(𝐶, [𝑀, 3𝑁]), and split the matrix and cipher-image is obtained. If plaintext image is gray, let
𝐶 into three equal-sized small matrixes as R , G and B 𝐶 = reshape(𝐶, [𝑀, 𝑁]), and obtain the cipher-image.
components. Combine R, G and B components into an image, The image encryption scheme is shown by Figure 9.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

B. Decryption Algorithm Step 4. If cipher-image is color, split the matrix 𝑆 into three
The decryption procedure is a reverse process of small matrixes with the same size as the three components of
encryption procedure. a color image, which are combined together, and a color
Step 1. We remove the diffusion effect from the last pixel to plaintext-image is obtained; If cipher-image is gray, 𝑆 is
the first pixel using the following formulae. equal to 𝑃, and gray plaintext-image is obtained.
𝑇i = 𝐶𝑖−1 ⊕ (𝐶𝑖 −𝑊 ̃𝑖 )mod 𝜃 if 𝑖 = 𝑀′ × 𝑁 ′
{𝑇𝑖 = 𝑇𝑖+1 ⊕ 𝐶i−1 ⊕ 𝐶𝑖 ⊕ 𝑊 ̃𝑖 if 1 < 𝑖 < 𝑀′ × 𝑁 ′ (22) IV. Experimental results
𝑇i = 𝑇𝑀×3𝑁 ⊕ (𝐶𝑖 −𝑊 ̃𝑖 )mod 𝜃 if 𝑖 = 1 A good image encryption algorithm should encrypt
where 𝑀′ × 𝑁 ′ is the size of of cipher-image 𝐶. different kinds of image into unrecognized cipher-image. As
Step 2. Eliminate the column confusion effect as follow: shown in Figure 10 (b), the four types of plaintext-image
𝑇1 {𝑉𝑖,𝑗 , 𝑗} ← 𝑇{𝑖, 𝑗} (23) have different properties of histograms, but their cipher-
images are all random-like. As shown in Figure 10 (d), the
where 𝑖 = 1,2, ⋯ , 𝑀′ and 𝑗 = 1,2, ⋯ , 𝑁 ′ .
distribution of pixel values of cipher-images is harmonious,
Step 3. Eliminate the row confusion effect as follow:
and attracters can’t obtain useful information from
𝑆{𝑖, 𝑈𝑖,𝑗 } ← 𝑇1 {𝑖, 𝑗} (24)
histograms of cipher-images.

(a) (b) (c) (d)

−𝟏𝟒
FIGURE 11. Key sensitivity tests in encryption. (a) plaintext-image of Lena (b) cipher-image 𝐂𝟏 with 𝐝𝟏 + 𝟏𝟎 ; (c) cipher-image 𝐂𝟐 with 𝐝𝟐 − 𝟏𝟎−𝟏𝟒 ; (d)
the difference between 𝐂𝟏 and 𝐂𝟐 .

(a) (b) (c) (d) (e)

FIGURE 12. Key sensitivity tests in decryption. (a) cipher-image (b) plaintext-image of Lena (c) decrypted image 𝐃𝟏 with 𝐝𝟏 + 𝟏𝟎−𝟏𝟒 ; (d) decrypted
image 𝐃𝟐 with 𝐝𝟐 − 𝟏𝟎−𝟏𝟒 ; (d) the difference between 𝐃𝟏 and 𝐃𝟐 .

V. Security analysis A sensitivity test was performed based on the color

plaintext-image of “Lena” and the results are shown in
A. Key security Figure 11 and 12. When 𝛫 = 5, through Eq. (11), we can
obtain five keys as 𝑑1 = 0.799826060252032 , 𝑑2 =
8

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

0.89908888703445 , 𝑑3 = 2.799938279444270 , 𝑑4 = VI. Statistical analysis

3.800186667046599 , 𝑑5 = 4.500150065664913 . As
shown in Figure 11 (d), the cipher-images are completely A. Correlation coefficient
different even if the change in the keys is very slight. As In general, there are strong correlations between adjacent
Figure 12 (b) demonstrates, the decrypted image was exactly pixels in a plaintext image, the correlations between two
the same as the original plaintext image with the correct keys adjacent pixels need to be decreased in the ciphered image to
𝑑𝑘 , where 𝑘 = 1, 2, ⋯ , 5. The decrypted images are random- resist statistical attacks. The correlations between the
like although the change of keys is slight as shown in Figure horizontal, vertical, and diagonal adjacent pixels can be
12 (c), (d) and (e). calculated by 3000 randomly selected pairs of adjacent
pixels from the plaintext image and encrypted image as
B. Key space follows [37]:
cov(𝛼,𝛽)
The key space should be large enough to resist exhaustive 𝜆𝑥𝑦 = (25)
√𝜑(𝛼)√𝜑(𝛽)
attacks. According to Figure 11 and 12, the operational
where
precision of 𝑑𝑘 was fixed as 10−14 . When we considered the 𝑁
precision of 𝑑𝑘 , the key space size was 1014𝛫 (𝛫 = 5), it is 1
cov(𝛼, 𝛽) = ∑(𝛼𝑖 − Φ(𝛼)) (𝛽𝑖 − Φ(𝛽))
approximately 2232 and much larger than the security 𝑁
𝑖=1
requirement for a key space of 2100 [36]. It’s obvious that, 𝑁 𝑁
brute-force attacks on our encryption algorithm is infeasible. 1 2 1
𝜑(𝛼) = ∑(𝛼𝑖 − Φ(𝛼)) , Φ(𝛼) = ∑ 𝛼𝑖
Furthermore, the key space can be enlarged greatly by 𝑁 𝑁
𝑖=1 𝑖=1
increase of 𝛫.

TABLE 2
CORRELATIONS BETWEEN ADJACENT PIXELS IN THE “LENA” PLAINTEXT AND CIPHERTEXTS
Direction Plaintext Ciphertext Ref. [38] Ref. [40] Ref. [41] Ref. [42] Ref. [43]
Horizontal 0.9765 0.0032 0.0013 −0.0047 −0.0226 −0.0001 0.0019
Vertical 0.9606 −0.0004 0.0006 0.0015 0.0041 0.0089 0.0012
Diagonal 0.9356 0.0059 0.0019 0.0030 0.0368 0.0091 0.0009

(a) (b) (c)

(d) (e) (f)

FIGURE 13. AC between 𝐑, 𝐆 and 𝐁 components of the plain color image Lena, (a) correlation between row pixels of 𝐑 and 𝐆 components of the
plaintext-image; (b) correlation between row pixels of 𝐆 and 𝐁 components of the plaintext-image; (c) correlation between row pixels of 𝐑 and 𝐁
components of the plaintext-image; (d) correlation between column pixels of 𝐑 and 𝐆 components of the plaintext-image; (e) correlation between
column pixels of 𝐆 and 𝐁 components of the plaintext-image; (f) correlation between column pixels of 𝐑 and 𝐁 components of the plaintext-image.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

(a) (b) (c)

(d) (e) (f)

FIGURE 14. AC between 𝐑, 𝐆 and 𝐁 components of the cipher-image, (a) correlation between row pixels of 𝐑 and 𝐆 components of the cipher-image;
(b) correlation between row pixels of 𝐆 and 𝐁 components of the cipher-image; (c) correlation between row pixels of 𝐑 and 𝐁 components of the
cipher-image; (d) correlation between column pixels of 𝐑 and 𝐆 components of the cipher-image; (e) correlation between column pixels of 𝐆 and 𝐁
components of the cipher-image; (f) correlation between column pixels of 𝐑 and 𝐁 components of the cipher-image.

The correlations between adjacent pixels in the plaintext in Table 3. The experimental results showed that the
image and encrypted image are shown in Table 2, which information entropies of the encrypted images were close to
demonstrates that the correlations of ciphered image were 8.
much smaller than the plain image of “Lena”. TABLE 3
THE COMPARISON OF INFORMATION ENTROPY
The correlation between R, G, B components of an image
can be measured intuitively and amply by AC [38]. Similar Text images Entropy
to Eq. (8), it is defined as: Lena (256 × 256) 7.9991
𝐸[(𝑥𝑖 −𝐸[𝑋𝑖 ])−(𝑥𝑖+1 −𝐸[𝑋𝑖 ])] Pepper (256 × 256) 7.9991
𝐴𝐶 = ])2 ]
(26)
𝐸[(𝑥𝑖 −𝐸[𝑋𝑖 Baboon (256 × 256) 7.9991
where 𝑋𝑖 and 𝑋𝑖+1 is a pixel sequence and another pixel Ref. [40] (Lena) 7.9991
sequence from different components respectively, 𝐸[∙] is the Ref. [41] (Lena) 7.9973
Ref. [42] (Lena) 7.9973
mathematical expectation. Ref. [43] (Lena) 7.9912
Figure 13 and 14 depicts the correlations between R, G, B Ref. [44] (Lena) 7.9963
components of the plain color image Lena and its cipher-
image respectively. It’s obvious that, the correlation values
between R, G, B components of cipher-image is much C. Difference attack analysis
smaller than plaintext-image, it means that the proposed An excellent image encryption tactics should differential
encryption algorithm can mask the correlations between R, attacks sensitively, which can be measured based on NPCR
G, B components effectively. and NPCR as follows [39]:
∑𝑖𝑗 𝜔(𝑖,𝑗)
NPCR = × 100% (28)
𝐹×𝐺
B. Information Entropy 1 𝐸(𝑖,𝑗)−𝐸 ′ (𝑖,𝑗)|
Randomness of an image is measured by information UACI = [∑𝑖,𝑗 × 100% (29) ]
𝐹×𝐺 255
entropy. Similar to Eq. (7), it can be calculated as follows: where the width and height of the image is represented by 𝐹
2𝑁 −1 and 𝐺 respectively, 𝐸(𝑖, 𝑗) and 𝐸 ′ (𝑖, 𝑗) are the 𝑖 th row
1 and 𝑗th pixel values from ciphered images before and after
𝛨(𝑠) = ∑ 𝜈(𝑠𝑖 )log 2 (27)
𝜈(𝑠𝑖 ) one pixel of the plain image been changed respectively, and
𝑖=0
where 𝜈(𝑠𝑖 ) is the probability of symbol 𝑠𝑖 and 𝑁 is the bit 𝜔(𝑖, 𝑗) is determined as
depth of the image. The information entropies for the 0 𝐸(𝑖, 𝑗) = 𝐸 ′ (𝑖, 𝑗)
𝜔(𝑖, 𝑗) = { (30)
encrypted “Lena”, “Pepper” and “Baboon” images are listed 1 𝐸(𝑖, 𝑗) ≠ 𝐸 ′ (𝑖, 𝑗)

10

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

TABLE 4 values of NPCR and UACI are as NPCR=99.6094% ,

NPCR AND UACI VALUES FOR THE ENCRYPTED IMAGES
UACI=33.4635% [37].
Text images NPCR(%) UACI(%)
Lena(256 × 256) 99.6246 33.5118 VII. Robustness analysis
Pepper(256 × 256) 99.6048 33.3828 Figure 15 denotes the experimental results of data loss
Baboon(256 × 256) 99.5885 33.4590
Ref. [40] (Lena) 99.6253 33.4807
attack to the encrypted images. The quality of the decrypted
Ref. [41] (Lena) 99.6100 33.5300 images decreases as the data loss size increases as shown in
Ref. [42] (Lena) 99.6096 33.4574 Figure 15 (e)-(g). As shown in Figure 15 (h), the main
Ref. [43] (Lena) 100 33.47 information of the plaintext images can be recognized
Ref. [44] (Lena) 99.6228 33.7041 despite of some date modification, because there are
correlations between loss data and reserved data.
The keys of our proposed encryption scheme are based on
In order to test the resistance of noise attack, we add salted
the plaintext image, so even if only one pixel of plaintext
pepper noise to the encrypted images. As the Figure 16
image is changed, the change in the ciphered image is
shown, quality of the decrypted images is degraded with the
significant, which can make differential attacks ineffective.
increase of noise intensity, but the decrypted images still be
Table 4 shows the NPCR and UACI results for “Lena”,
identified. Therefore, the proposed encryption scheme can
“Pepper” and “Baboon” where all the NPCR and UACI
highly robust against noise and loss attacks.
scores are close to the expected values, where the expected

(a) (b) (c) (d)

(e) (f) (g) (h)

FIGURE 15. Experimental results for cipher-image of Lena with different kinds of data loss. (a) 𝟏/𝟏𝟔 data loss; (b) 𝟑/𝟏𝟔 data loss; (c) 𝟏/𝟒 data loss (d)
𝟏/𝟏𝟔 data modification with a square; (e), (f), (g) and (h) are the corresponding decrypted images.

(a) (b) (c)

11

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

(d) (e) (f)

FIGURE 16. Experimental results for a noise attack. (a) 0.01 intensity salted pepper noise; (b) 0.05 intensity salted pepper noise; (c) 0.1 intensity salted
pepper noise; (d), (e) and (f) are the corresponding decrypted images.

VIII. Conclusion [12] Wang, X. Y., Li, P., Zhang, Y. Q., et al. A novel color image encryption
In this study, we have proposed an image encryption scheme using DNA permutation based on the Lorenz system,
Multimedia Tools and Applications, 77 (5) 6243-6265 2018.
scheme based on improved Lorenz system. The proposed [13] Rostami, M. J., Shahba, A., Saryazdi, S., et al. A novel parallel image
system has more complex kinetic properties than Lorenz encryption with chaotic windows based on logistic map, Computers &
system, therefore it is more suitable to encrypt image. In Electrical Engineering, 62 384-400 2017.
encryption algorithm, the keys are associated with plaintext- [14] Bansal, R., Gupta, S., Sharma, G. An innovative image encryption
scheme based on chaotic map and Vigenere scheme, Multimedia Tools
image to update the keys with the change of plaintext image, and Applications, 76 (15) 16529-16562 2017.
and key space is large enough to resist attacks of brute-force. [15] Kanafchian, M., Fathi-Vajargah, B. A novel image encryption scheme
Through the analysis of key security, protection against based on clifford attractor and noisy logistic map for secure
differential attack, information entropy and correlation, we transferring images in navy, International Journal of E-Navigation and
Maritime Economy, 6, 53-63 2017.
can see that our proposed system can encrypt different kinds [16] Liu, H. J., Kadir, A., Niu, Y. J. Chaos-based color image block
of images into random-like cipher-images with high security encryption scheme using S-box, AEU-International Journal of
level. Electronics and Communications, 68 (7) 676-686 2014.
[17] Chen, J. X., Zhu, Z. L., Fu, C., et al. A fast chaos-based image
encryption scheme with a dynamic state variables selection mechanism,
REFERENCES Communications in Nonlinear Science and Numerical Simulation 20
[1] Lorenz, E. Deterministic nonperiodic flow, Journal of the Atmospheric
(3) 846-860 2015.
Sciences, 20 (2) 130–141 1963.
[18] Guesmi, R., Ben Farah, M. A., Kachouri, A., et al. Hash key-based
[2] Nazari, S., Ataei, M., Tamizi, A. Improving diagnosis of some brain
image encryption using crossover operator and chaos, Multimedia
disease by analysing chaotic indices of EEG signals, International
Tools and Applications, 75 (8) 4753-4769 2016.
Journal of Biomedical Engineering and Technology, 22 (4) 349-369
[19] Song, C. Y., Qiao, Y. L. A novel image encryption algorithm based on
2016.
DNA encoding and spatiotemporal chaos, Entropy, 17 (10) 6954-
[3] Zare, K., Chesley, S. Order and chaos in the planar isosceles three-body
6968 2015.
problem, Chaos 8 (2) 475-494 1998.
[20] Wang, X. Y., Feng, L., Zhao, H. Y. Fast image encryption algorithm
[4] Chen, M. Y. Chaos synchronization in complex networks, IEEE
based on parallel computing system, Information Sciences, 486 340-
Transactions on Circuits and Systems I-Regular Papers, 55 (5) 1335-
358 2019.
1346 2008.
[21] Al-Hazaimeh, O. M., Al-Jamal, M. F., Alhindawi, N., et al. Image
[5] Gong, Y. B., Xu, B., Xu, Q., et al. Ordering spatiotemporal chaos in
encryption algorithm based on Lorenz chaotic map with dynamic secret
complex thermosensitive neuron networks, Physical Review E, 73 (2)
keys, Neural Computing & Applications, 31 (7) 2395-2405 2019.
2006.
[22] Anees, A. An image encryption scheme based on Lorenz system for
[6] Hommes, C. H., Nusse, H. E., Simonovits, A. Cycles and chaos in a
low profile applications, 3D Research, 6 (3) 2015.
socialist economy, Journal of Economic Dynamics & Control, 19 (1-2)
[23] Younas, I., Khan, M. A new efficient digital image encryption based
155-179 1995.
on inverse left almost semi group and Lorenz chaotic system, Entropy
[7] Ge, X., Lu, B., Liu, F. L., Luo, X. Y. Cryptanalyzing an image
20 (12) 2018.
encryption algorithm with compound chaotic stream cipher based on
[24] Hussain, I., Anees, A., Alkhaldi, A. H., et al. Construction of chaotic
perturbation, Nonlinear Dynamics, 90 (2) 1141-1150 2017.
quantum magnets and matrix Lorenz systems S-boxes and their
[8] Pak, C., Huang, L. L. A new color image encryption using combination
applications, Chinese Journal of Physics, 56 (4) 1609-1621 2018.
of the 1D chaotic map, Signal Processing, 138 129-137 2017.
[25] Girdhar, A., Kumar, V. A RGB image encryption technique using
[9] Zhou, J. T., Au, O. C. On the Security of Chaotic Convolutional Coder,
Lorenz and Rossler chaotic system on DNA sequences, Multimedia
IEEE Transactions on Circuits and Systems I-Regular Papers, 58 (3)
Tools and Applications, 77 (20) 27017-27039 2018.
595-606 2011.
[26] Kaur, M., Kumar, V. Efficient image encryption method based on
[10] Liu, H., Wan, H. B., Tse, C. K., et al. An encryption scheme based on
improved Lorenz chaotic system, Electronics Letters, 54 (9) 562-563
synchronization of two-layered complex dynamical networks. IEEE
2018.
Transactions on Circuits and Systems I-Regular Papers, 63 (11) 2010-
[27] Arshad, U., Batool, S. I., Amin, M. A novel image encryption scheme
2021 2016.
based on walsh compressed quantum spinning chaotic Lorenz system,
[11] Li, K. Z., Zhao, M. C., Fu, X. C. Projective synchronization of driving-
International Journal of Theoretical Physics, 58(10) 3565-3588 2019.
response stems and its application to secure communication, IEEE
[28] Bouslehi, H., Seddik, H. Innovative image encryption scheme based on
Transactions on Circuits and Systems I-Regular Papers, 56 (10) 2280-
a new rapid hyperchaotic system and random iterative permutation,
2291 2009.
Multimedia Tools and Applications, 77 (23) 30841-30863 2018.

12

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 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.2020.2988880, IEEE Access

[29] Li, Z., Peng, C. G., Li, L. R., et al. A novel plaintext-related image
encryption scheme using hyper-chaotic system, Nonlinear Dynamics, QIANG ZHANG is a professor at Dalian
94 (2) 1319-1333 2018. University of Technology, Dalian, China. He is
[30] Ran, Q. W., Wang, L., Ma, J., et al. A quantum color image encryption also a professor at Dalian University, Dalian,
scheme based on coupled hyper-chaotic Lorenz system with three China. He has received his Ph.D. degree in 2002 at
impulse injections, Quantum Information Processing, 17 (8) 2018. Xidian University, Xi’an China. His research
[31] Pincus, S. M., Approximate entropy as a measure of system-complexity, interests are intelligent computing and intelligent
Proceedings of the National Academy of Science of the United States robots. Now he has served as editorial boards of
of America, 88 (6) 2297-2301 1991. seven international journals and has edited special
[32] Shannon, C. E. A mathematical theory of communication, Bell System issues in journals such as Neurocomputing and
Technical Journal, 27 (4) 623-656 1948. International Journal of Computer Applications in
[33] Chen, C., Sun, K. H., Peng, Y. X., et al. A novel control method to Technology.
counteract the dynamical degradation of a digital chaotic sequence, XIAOPENG WEI is a professor at Dalian
European Physical Journal Plus, 134 (1) 2019. University of Technology, Dalian, China. His
[34] Fan, C. L., Ding, Q., Tse, C. K. Counteracting the dynamical research area includes computer animation,
degradation of digital chaos by applying stochastic jump of chaotic intelligent CAD. So far, he has (co-) authored
orbits, International Journal of Bifurcation and Chaos, 29 (8) 2019. about 160 papers published.
[35] Wang, X. Y., Wang, Y., Zhu, X. Q., et al. A novel chaotic algorithm
for image encryption utilizing one-time pad based on pixel level and
DNA level, Optics and Lasers in Engineering, 125 2020.
[36] Wang, X. Y., Zhu, X. Q., Wu, X. J., et al. Image encryption algorithm
based on multiple mixed hash functions and cyclic shift, Optics and
Lasers in Engineering, 107 370-379 2018.
[37] Gan, Z. H., Chai, X. L., Han, D. J., et al. A chaotic image encryption
algorithm based on 3-D bit-plane permutation, Neural Computing & CHANJUAN LIU received the B.Eng. degree in
Applications, 31 (11): 7111-7130 2019. computer science from Jinan University,
[38] Hua, Z. Y., Zhou, Y. C. Image encryption using 2D Logistic-adjusted- Guangzhou, China, in 2010, and the Ph.D. degree
Sine map, Information Sciences, 339 237-253 2016. in computer science from Peking University,
[39] Chai, X. L., Gan, Z. H., Yang, K., et al. An image encryption algorithm Beijing, China, in 2016. She is currently a
based on the memristive hyperchaotic system, cellular automata and Lecturer with the Dalian University of
DNA sequence operations, Signal Processing-Image Communication, Technology, Dalian, China. Her research interests
52 6-19 2017. include logic and game theory.
[40] Zahmoul, R., Ejbali, R., Zaied, M. Image encryption based on new Beta
chaotic maps, Optics and Lasers in Engineering, 96 39-49 2017.
[41] Xu, L., Gou, X., Li, Z., et al. A novel chaotic image encryption
algorithm using block scrambling and dynamic index based diffusion,
Optics and Lasers in Engineering, 91 41-52 2017.
[42] Cao, C., Sun, K. H., Liu, W. H. A novel bit-level image encryption
algorithm based on 2D-LICM hyperchaotic map, Signal Processing,
143 122-133 2018.
[43] Wu, J. H., Liao, X. F., Yang, B. Color image encryption based on
chaotic systems and elliptic curve ElGamal scheme, Signal Processing,
141 109-124 2017.
[44] Belazi, A., Abd El-Latif, A. A., Belghith, S. A novel image encryption
scheme based on substitution-permutation network and chaos, Signal
Processing, 128 155-170 2016.

CHENGYE ZOU received the B.S. degree in

physics, the M.E. degree in theoretical physics
from the Liaoning Normal University, Dalian,
China, and Ph.D. degree in computer application
technology from Dalian University of Technology,
Dalian, China. He is currently a Lecturer with the
Anyang Normal University, Anyang, China. His
research interests include complex networks,
biological computing and chaotic image
encryption.

13

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.