Hindawi

Security and Communication Networks

Volume 2019, Article ID 8694678, 15 pages
https://doi.org/10.1155/2019/8694678

Research Article
Image Encryption with Double Spiral Scans and Chaotic Maps

Zhenjun Tang ,1 Ye Yang,1 Shijie Xu,1 Chunqiang Yu ,2 and Xianquan Zhang 1,2

1
Guangxi Key Lab of Multi-Source Information Mining & Security, and Department of Computer Science,
Guangxi Normal University, Guilin 541004, China
2
Network Information Center, Guangxi Normal University, Guilin, 541004, China

Correspondence should be addressed to Zhenjun Tang; [email protected]

Received 19 August 2018; Revised 9 December 2018; Accepted 24 December 2018; Published 15 January 2019

Academic Editor: Angel M. Del Rey

Copyright © 2019 Zhenjun Tang et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Image encryption is a useful technique of image content protection. In this paper, we propose a novel image encryption algorithm
by jointly exploiting random overlapping block partition, double spiral scans, Henon chaotic map, and Lü chaotic map. Specifically,
the input image is first divided into overlapping blocks and pixels of every block are scrambled via double spiral scans. During
spiral scans, the start-point is randomly selected under the control of Henon chaotic map. Next, image content based secret keys
are generated and used to control the Lü chaotic map for calculating a secret matrix with the same size of input image. Finally, the
encrypted image is obtained by calculating XOR operation between the corresponding elements of the scrambled image and the
secret matrix. Experimental result shows that the proposed algorithm has good encrypted results and outperforms some popular
encryption algorithms.

1. Introduction transform and a chaotic map to generate secure matrix for

block-wise encryption. In [6], Zhang and Liu used skew-tent
Some well-known security events, such as PRISM and chaotic map to achieve permutation and diffusion without
Xkeyscore, make people pay much attention to information changing pixel information. This method has high efficiency
security. Since digital images are widely used in the Internet,
and a large key space, but it is not secure enough from the
how to protect image content [1, 2] has become an issue to
be urgently solved. Image encryption is a useful technique viewpoint of histogram [5]. In [7], Li et al. exploited reversible
of image content protection [3]. It converts images into data hiding (RDH) and compressive sensing to design a
noise-like encrypted images by disrupting pixel positions meaningful image encryption algorithm. This algorithm
or changing pixel values. In recent years, researchers have encrypts a secret image into a meaningful image by RDH
developed many useful image encryption algorithms. These and reaches a high embedding rate. Recently, Wang et al.
algorithms can be roughly divided into two categories as [8] designed an efficient image encryption algorithm based
follows. on two-dimensional partitioned cellular automaton. This
The first direction is to manipulate image pixels in the algorithm supports parallel computing and is easy for VLSI
spatial domain. This kind of algorithm usually scrambles implementation. In another work, Wang et al. [9] exploited
pixel positions through matrix transformation and destroys multiple mixed hash functions, cyclic-shift function, and
spatial correlation between pixels of the original image, so piece-wise linear chaotic maps to achieve image encryption.
as to convert input image into chaotic image. For example, This scheme can overcome security flaw of the well-known
Tang et al. [4] proposed an encryption algorithm based chaotic image encryption called Baptista’s algorithm and
on Arnold transform and three random strategies. This its improved versions. In [10], Hayat and Azam proposed
encryption technique is a secure algorithm and can overcome a useful image encryption technique using a dynamic S-
size limitation of the Arnold transform. In another study, box and pseudo-random numbers over an elliptic curve.
Tang et al. [5] divided input image into overlapping blocks, This technique can resist known plaintext attack and chosen
conducted random block shuffling, and exploited Arnold plaintext attack.
2 Security and Communication Networks

Since chaotic systems have many sensitive properties Hénon-Sine map. This algorithm can resist statistical attack,
(e.g., sensitivity to initial conditions and system parameters) differential attack, and noise attack, but has limitation in
and show better performance than traditional encryption encrypting color images.
techniques (e.g., AES [11] and DES [12]), many researchers The other direction is to conduct encryption in the trans-
have tried to design image encryption with chaotic maps form domain [35, 36]. Generally, this kind of encryption algo-
[13–19]. In general, chaos-based image encryption algorithms rithms firstly transforms input image from spatial domain
consist of two steps: pixel permutation and pixel diffusion. to transform domain, then modifies those coefficients in
The pixel permutation changes pixel position, while the pixel the transform domain with some well-defined rules, and
diffusion alters pixel values where a change in a pixel will finally converts coefficients to spatial domain. For example,
spread almost to other pixels of entire image. Contributed Singh et al. [35] exploited Arnold transform and singular
by the sensitivity properties of chaotic system, chaos-based value decomposition to make phase image encryption in the
image encryption algorithms generally achieve good security
fractional Hartley domain. Vashisth et al. [36] conducted
performance. Some representative chaos-based encryption
image encryption in the fractional Mellin transform domain
algorithms are introduced here. Amin et al. [20] proposed
a new image encryption algorithm based on chaotic block by using structured phase filters and phase retrieval. Naeem
cipher. This algorithm jointly uses cryptographic primitive et al. [37] presented novel image encryption algorithms with
operations, nonlinear transformation functions, and chaotic a cyclic shift and the 2D chaotic Baker map in transform
tent map to achieve encryption. It is secure against brute- domains, such as the Integer Wavelet Transform (IWT)
force attack. Abd El-Latif et al. [21] presented a novel image domain, the Discrete Wavelet Transform (DWT) domain,
encryption with linear feedback shift register and chaotic and the Discrete Cosine Transform (DCT) domain. The
maps in time and frequency domains. This method can resist algorithm in DWT domain shows better performance than
differential attack. In another study, Abd El-Latif and Niu those in other transform domains. In another work, Belazia et
[22] proposed a hybrid image encryption by using chaotic al. [38] proposed a novel partial image encryption approach
system and cyclic elliptic curve. This method reaches good based on permutation-substitution-diffusion (PSD) network
security. In [23], Tang et al. exploited Henon map, logistic and multiple chaotic maps in wavelet transform domain.
map, and bit-plane decomposition to design an algorithm for Recently, Annaby et al. [39] integrated random fractional
multiple-image encryption. This algorithm can convert four Fourier transforms, phase retrieval, and chaotic maps to
gray-scale images into an encrypted PNG image. In another design a scheme for color image encryption. Zhang and Tong
work, Wang et al. [24] used two chaotic systems to develop [40] exploited IWT and set partitioning in hierarchical trees
a hybrid color image encryption scheme. In [25], Abanda (SPIHT) to make image encryption and compression. This
and Tiedeu proposed a fast and simple image encryption method has good performances in security and compression.
algorithm by combining two kinds of chaotic maps to meet Li et al. [41] exploited two-dimensional DWT to decompose
real-time application. Belazi et al. [26] presented a novel original images and conducted encryption in DWT domain
selective image encryption by using DWT with AES s-box by Arnold transform and robust chaotic map. This algorithm
and chaotic permutation. This method can resist differential can convert four grayscale images with the same size into an
and statistical attacks. In another study, Belazi et al. [27] encrypted image. Wu et al. [42] introduced an asymmetric
designed an efficient image encryption with substitution- multiple-image encryption method via compressed sensing
permutation network and chaotic systems. This algorithm has and nonlinear operations in cylindrical diffraction domain.
good performances in security and speed. This method can encrypt eight images and resist ciphertext-
Recently, Tang et al. [28] proposed an image encryption only attack.
algorithm by using random projection partition and chaotic In this paper, we propose an image encryption algorithm
system. This algorithm is secure and has a fast speed. Li based on double spiral scans and chaotic maps. The pro-
et al. [29] introduced a quantum color image encryption posed algorithm jointly exploits random overlapping block
scheme. This scheme exploits quantum controlled-NOT partition, double spiral scans, Henon chaotic map, and Lü
image generated by multiple chaotic maps to control the XOR chaotic map to calculate encrypted image. Compared with
operation in the encryption process. It can resist the attack previous image encryption algorithms based on chaotic map,
of histogram analysis. Parvaz and Zarebnia [30] defined a key contribution of our algorithm is the double spiral scans,
a combination chaotic system with logistic, sine, and tent which can efficiently scramble pixels of image block. Many
systems and applied it to image encryption. To improve experiments are conducted and the results show that the pro-
security, Liu et al. [31] proposed to use a randomly sampled posed algorithm is effective and outperforms some popular
noise signal as initial value of chaotic map. Chen and Hu encryption algorithms. The rest of this paper is organized as
[32] designed an adaptive encryption algorithm based on follows. In Section 2, we introduce the proposed algorithm.
improved chaotic mapping for medical images. However, the In Section 3, we present key space analysis. In Section 4,
encryption results of this algorithm have obvious block effect. we discuss experimental results. Finally, conclusions of this
In [33], Chai et al. jointly used chaotic system, elementary paper are made in Section 5.
cellular automata, and compressive sensing to design efficient
image encryption. This algorithm can resist known-plaintext 2. Proposed Algorithm
attack and chosen-plaintext attack. Wu et al. [34] proposed
a new image encryption algorithm by pixel diffusion with Figure 1 shows the block diagram of our image encryption.
DNA approach and pixel permutation by a two-dimensional The main steps of our algorithm include random overlapping
Security and Communication Networks 3

keys Henon map

keys

Random
Input image overlapping block Double spiral scans XOR operation Encrypted image
partition

key k1

Key generation Lü map

Figure 1: Block diagram of our image encryption.

block partition, double spiral scans, and XOR operation. where ⌊⋅⌋ means rounding down operation and mod(⋅, ⋅)
In the first step, the input image is randomly divided into means modulo operation. Therefore, the total number of
overlapping blocks under the control of secret keys. In the random overlapping blocks is 𝑁total = 𝑛𝑥 ×𝑛𝑦 . The coordinate
second step, we shuffle image pixels of every overlapping of the i-th image block in the x-axis direction is denoted by
block by double spiral scans, where a secret key k1 is used to 𝑋[𝑖], and the coordinate of the j-th image block in the y-
determine the order of encrypting image blocks. To improve axis direction is denoted by 𝑌[𝑖], where 𝑖 = 1, 2, . . . , 𝑛𝑥 and
security, the Henon chaotic map is exploited to generate 𝑗 = 1, 2, . . . , 𝑛𝑦 . Then, 𝑋[𝑖] and 𝑌[𝑗] can be determined as
random start-points for double spiral scans. In the third step, follows. If mod(N − t 𝑥 , S − t 𝑥 ) = 0, 𝑋[𝑖] = (𝑖 − 1)(𝑆 − 𝑡𝑥 ),
we generate keys based on the content of input image and use where 𝑖 = 1, 2, . . . , 𝑛𝑥 . Otherwise, the x-coordinates of the
the keys to control the Lü chaotic map for generating a secret first n𝑥 − 1 blocks are calculated by X [i] = (i − 1)(S − t 𝑥 ), where
matrix. Finally, we calculate the XOR operation between 𝑖 = 1, 2, . . . , 𝑛𝑥 − 1, and the x-coordinate of the last block is
the shuffled image and the secret matrix, and the result X[n𝑥 ] = N − S + 1. Similarly, if mod(M − t 𝑦 , S − t𝑦 ) = 0, Y[j]
is the final encrypted image. In the following sections, we = (j − 1)(S − t 𝑦 ), where 𝑦 = 1, 2, . . . , 𝑛𝑦 . Otherwise, the y-
first introduce the random overlapping-block partition, then coordinates of the first n𝑦 − 1 blocks are Y[j] = (j − 1)(S − t 𝑦 ),
describe the double spiral scans and the used chaotic maps where 𝑦 = 1, 2, . . . , 𝑛𝑦 − 1. And the y-coordinate of the last
in our algorithm, and finally explain the key generation and block is Y[n𝑦 ] = M − S+1. Here, image blocks are numbered
the detailed steps of our encryption scheme and decryption from left to right and top to bottom, and the coordinates of
scheme. the i-th image block are represented by (X[u𝑖 ], Y[v𝑗 ]). The
random block pattern depends on the block size S and the
2.1. Random Overlapping Block Partition. We exploit random overlapping sizes 𝑡𝑥 and 𝑡𝑦 , where the theoretical range of S is
overlapping-block partition scheme [5] to divide input image (1, min (M, N)]. In the experiment, it is found that a small S
into overlapping blocks. And then we shuffle every image value will make more image blocks and thus lead to a slow
block by double spiral scans to complete the scrambling speed. Therefore, we randomly select the S value from the
operation. The detailed process of the random overlapping- range [32, min(𝑀, 𝑁)) in this study. As the ranges of t 𝑥 and t 𝑦
block partition is explained as follows. Assume that the size are both [1, 𝑆), the total number of our random block patterns
of input image is 𝑀 × 𝑁, the selected block size is 𝑆 × 𝑆, and is (min(𝑀, 𝑁) − 31)(𝑆 − 1)2 . Note that the parameters S, t 𝑥 ,
𝑡𝑥 and 𝑡𝑦 are the overlapping sizes between adjacent blocks and t 𝑦 are determined by the user and thus they can be taken
along the x-axis and the y-axis, respectively, where 𝑡𝑥 ∈ [1, 𝑆) as secret keys in practice.
and 𝑡𝑦 ∈ [1, 𝑆). Let 𝑛𝑥 and 𝑛𝑦 be the numbers of image blocks
in the x-axis and the y-axis, respectively. Thus, 𝑛𝑥 and 𝑛𝑦 can 2.2. Double Spiral Scans. The scheme of double spiral scans
be calculated as follows. proposed in this paper is used to scramble position of every
𝑁 − 𝑡𝑥 pixel in an overlapping block. Details of our double spiral
{
{ 𝑆−𝑡 ,
{ If mod (𝑁 − 𝑡𝑥 , 𝑆 − 𝑡𝑥 ) = 0 scans for pixel scrambling in a block are explained as follows.
𝑛𝑥 = { 𝑁 − 𝑡 𝑥 (1) As shown in Figure 2, our double spiral scans consist of
{
{⌊ 𝑥
⌋ + 1, Otherwise two parts. Firstly, a start-point is randomly selected. Then,
{ 𝑆 − 𝑡𝑥 we visit block pixels from the start-point with a spiral scan as
𝑀 − 𝑡𝑦 shown in Figure 2(a) and obtain a pixel sequence according
{
{ , If mod (𝑀 − 𝑡𝑦 , 𝑆 − 𝑡𝑦 ) = 0
{
{ 𝑆 − 𝑡𝑦
to the order of visit. Similarly, we visit block pixels from the
𝑛𝑦 = { 𝑀 − 𝑡 (2) start-point with another spiral scan as shown in Figure 2(b)
{
{ 𝑦 and obtain another pixel sequence according to the order of
{⌊ ⌋ + 1, Otherwise
{ 𝑆 − 𝑡𝑦 the visit. Next, we concatenate the first pixel sequence and the
4 Security and Communication Networks

(a) The first scanning direction (b) The second scanning direction

Figure 2: Diagram of double spiral scans.

A B C D E N O J E D Q R S T O
F G H I J C B A F K J E D C B
K L M N O P U V W X A F K P U
P Q R S T Y T S R Q V W X Y L
U V W X Y M L G H I G H I N M

(a) Original block (b) Encrypted block (c) Encrypted block

started from ‘N’ started from ‘Q’

Figure 3: Original block and different encrypted blocks.

Pixel sequence 1: A-B-C-D-E-F-G-H-I-J-K-L-M-N-O-P-Q-R-S-T-U-V-W-X-Y.

Pixel sequence 2: N-O-J-E-D-C-B-A-F-K-P-U-V-W-X-Y-T-S-R-Q-M-L-G-H-I.
Pixel sequence 3: Q-R-S-T-O-J-E-D-C-B-A-F-K-P-U-V-W-X-Y-L-G-H-I-N-M.

Figure 4: Original pixel sequence and the pixel sequences generated by double spiral scans with different start-points.

second pixel sequence to make a new pixel sequence. Finally, scanning direction. If there is no pixel for scanning, the
we can generate the encrypted image block by filling image second scanning process is also finished and then the second
block from left to right and top to bottom via picking pixel pixel sequence is obtained. Next, a new pixel sequence can
from the new sequence one by one. Note that all block pixels be generated by concatenating the first and the second pixel
are visited, and every pixel is scanned only once. sequences. Finally, the encrypted block can be obtained by
More specifically, we first randomly select start-point for filling it with the new pixel sequence from left to right and
the double spiral scans. Here, coordinates of the random top to bottom. Clearly, the encrypted block can be accurately
start-point are randomly generated by the Henon chaotic decrypted once the start-point is known by filling pixels back
map. The detailed calculation will be described in Section 2.3. according to the visiting order of double spiral scans.
As shown in Figure 2, the solid circle is the start-point of An example of our double spiral scans is illustrated here.
double spiral scans in a block. The scanning process can be Figure 3(a) is a 5×5 image block. We scan pixels from left
divided into two directions. For the first scanning direction, to right and top to bottom and then get a pixel sequence
as shown in Figure 2(a), we scan block pixels starting from 1, as shown in Figure 4. Suppose that coordinates of the
the start-point and follow the below scanning direction: right, start-point are (3, 4), i.e., the location of ‘N’ as shown in
up, left, and down. When scanning to block border or the Figure 3(a). We visit block pixels by the double spiral scans
scanned pixel, we turn the scanning direction. If there is and generate the pixel sequence 2, as shown in Figure 4. Then,
no pixel for scanning, the first scanning process is finished we fill pixels back to image block with the pixel sequence
and then the first pixel sequence is obtained. For the second 2 and obtain the encrypted block as shown in Figure 3(b).
scanning direction, as shown in Figure 2(b), we scan block Similarly, suppose that coordinates of the start-point are (5,
pixels starting from the start-point and follow the below 2), i.e., the location of ‘Q’ as shown in Figure 3(a). We visit
scanning direction: left, up, right, and down. Similarly, when block pixels by the double spiral scans and generate the pixel
scanning to block border or the scanned pixel, we turn the sequence 3, as shown in Figure 4. Then, we fill pixels back
Security and Communication Networks 5

to image block with the pixel sequence 3 and obtain the [(1/𝑀𝑁) ∑𝑁 𝑀
𝑖=1 ∑𝑗=1 𝐼 (𝑖, 𝑗)]
encrypted block as shown in Figure 3(c). Obviously, different 𝑦0 =
255
start-points lead to different encrypted results. In this study,
we choose different start-points for different image blocks by 𝑧0 = 𝑥0 + 𝑦0
using Henon chaotic map. This strategy can improve security (7)
of our algorithm.
where 𝑀 × 𝑁 is the size of input image and 𝐼(𝑖, 𝑗) is the pixel
2.3. Chaotic Maps. This section describes the chaotic maps value of input image I, where 𝑖 ∈ [1, 𝑁] and 𝑗 ∈ [1, 𝑀].
used in our algorithm. Henon chaotic map is a typical two- Obviously, x0 is the decimal result of XOR operation between
dimensional discrete chaotic map. We use it to generate the all image pixel values, and y0 is the decimal result of mean
start-points for double spiral scans. The Henon chaotic map value of all image pixels. Therefore, if an image pixel in the
is defined as follows. plaintext image is changed, the calculated results of x0 , y0 , and
z0 are also changed and then the chaotic sequences controlled
𝑥 (𝑘 + 1) = 1 − 𝑎𝑥2 (𝑘) + 𝑦 (𝑘) by these keys will be different. This means that a changed
(3) input image will lead to a different encrypted result.
𝑦 (𝑘 + 1) = 𝑏𝑥 (𝑘)

where a and b are control parameters. When 𝑎 ∈ (0.54, 2) and 2.5. Encryption Scheme. The steps of our encryption scheme
𝑏 ∈ (0, 1), the Henon chaotic map will reach chaotic state. are as follows.
In this study, we select 𝑎 = 1.4 and 𝑏 = 0.3 and take the
initial values x(0) and y(0) as keys. We repeatedly calculate Step 1. The input image I is divided into random overlapping
(3) 𝑁total times and then obtain two chaotic sequences: x = blocks according to the block size S and the overlapping
[𝑥(1), 𝑥(2), . . . , 𝑥(𝑁total )] and y = [𝑦(1), 𝑦(2), . . . , 𝑦(𝑁total )]. sizes t 𝑥 and t 𝑦 . Calculate the total number of the overlapping
Since elements of the chaotic sequences are decimals and blocks 𝑁total and use a pseudo-random generator controlled
pixel coordinates of image block are integers, the two by a key k1 to generate 𝑁total random numbers. Sort these
sequences are mapped to integer sequences as follows. random numbers and record the original positions of the
sorted numbers in an array P[𝑁total ], which is used to
𝐷 (𝑖) = ⌊(mod (𝑥 (𝑖) ∗ 248 , 𝑆))⌋ , 𝑖 = 1, 2, . . . , 𝑁total (4) determine the order of encrypting image blocks.

𝐹 (𝑖) = ⌊(mod (𝑦 (𝑖) ∗ 248 , 𝑆))⌋ , 𝑖 = 1, 2, . . . , 𝑁total (5) Step 2. Use Henon chaotic map to generate two arrays:
𝐷[𝑁total ] and 𝐹[𝑁total ]. Note that D(i) and F(i) are the x-
where D(i) is the i-th element of the array D used to record coordinate and y-coordinate of the start-point of the P[i]-th
the x-coordinate of the start-point of the i-th image block and block, respectively, where 1 ≤ 𝑖 ≤ 𝑁total . Take (𝐷(𝑖), 𝐹(𝑖)) as
F(i) is the i-th element of the array F used to record the y- the start-point of the P[i]-th block, encrypt the P[i]-th block
coordinate of the start-point of the i-th image block. by double spiral scans, and write the encrypted result to the
Moreover, we exploit Lü chaotic map to generate secret image. Repeatedly conduct block encryption starting from i
matrix for XOR operation. The classical Lü chaotic map [43] = 1 to i = 𝑁total . After all blocks are processed, a scrambled
is a three-dimensional discrete chaotic map that characterizes image J is available.
the transition between the Lorenz system [44] and the Chen
system. The Lü chaotic map is calculated as follows. Step 3. Calculate content-based keys x0 , y0 , and z0 from input
image and use them to control the Lü chaotic map to generate
𝑥󸀠 = 𝑎 (𝑦 − 𝑥) a secret matrix G sized M×N. Then calculate E = J ⊕ G,
where ⊕ represents the XOR operation of the corresponding
𝑦󸀠 = 𝑐𝑦 − 𝑥𝑧 (6) elements of the matrices, and the matrix E is the encrypted
image.
𝑧󸀠 = 𝑥𝑦 − 𝑏𝑧
2.6. Decryption Scheme. The decryption scheme is a reverse
where x0 , y0 , and z0 are initial values of the Lü chaotic system process of our encryption scheme. Note that the keys of
and a, b, and c are its control parameters. The system is in the Lü system are transmitted to the receiver via secure
chaos when a = 36, b = 3, and c = 20. Note that x0 , y0 , and z0 channel. Therefore, calculations of initial chaotic parameters
are also taken as keys. x0 , y0 , and z0 are not required at the receiver’s side. Detailed
decryption process is as follows.
2.4. Key Generation. We generate content-based keys based
on input image and use them to control the Lü chaotic map. Step 1. Use x0 , y0 , and z0 to control the Lü chaotic map to
This strategy can make our algorithm resistant to differential generate a secret matrix G. Calculate J = E ⊕ G.
attack. To do so, the initial values x0 , y0 , and z0 are calculated
by the following equations. Step 2. According to the block size S and the overlapping
sizes t 𝑥 and t 𝑦 , the image matrix J is divided into random
[𝐼 (1, 1) ⊕ 𝐼 (1, 2) ⊕ ⋅ ⋅ ⋅ ⊕ 𝐼 (𝑖, 𝑗)] overlapping blocks. Compute the number of the overlapping
𝑥0 = blocks 𝑁total and use a pseudo-random generator controlled
255
6 Security and Communication Networks

(a) Lena (b) Fingerprint (c) Woman (d) Lake

(e) Goldhill (f) ChestXray

Figure 5: Six original images.

by a key k1 to generate 𝑁total random numbers. Sort these used to randomly select block and the permutation number of
random numbers and record the original positions of the blocks is 𝑁total !, the valid key space is min(264 , 𝑁total !). When
sorted numbers in an array 𝑃[𝑁total ]. the block number 𝑁total ≥ 21, the following expression 𝑁total !
> 264 holds. In this case, the key space is 264 . Otherwise, the
Step 3. Use Henon map to calculate the arrays 𝐷[𝑁total ] and
key space is 𝑁total !. The third part is the initial values of the Lü
𝐹[𝑁total ]. Take (𝐷(𝑖), 𝐹(𝑖)) as the start-point of the P[i]-th
chaotic map. These three parameters are all floating numbers.
block, decrypt the P[i]-th block by double spiral scans, and
write the decrypted result to the image. Repeatedly conduct Therefore, the key space of this part is 264×3 = 2192 .
block decryption starting from 𝑖 = 𝑁total to 𝑖 = 1. After In summary, the key space of our algorithm is 2128 ×
all blocks are processed, the decrypted image I is obtained. (min(M, N) − 31)(S − 1)2 × min(264 , 𝑁total !) × 2192 = (min(M,
Note that this step is similar to the second step of encryption N) − 31)(S − 1)2 × min(264 , 𝑁total !) × 2320 . For example, for
scheme. The major difference is the order of processing image a 512 × 512 image, if S = 150, t 𝑥 = 70, and t𝑦 = 70, the total
blocks. number of image blocks is 36. Consequently, our key space is
481 × 1492 × 264 × 2320 = 481 × 1492 × 2384 ≈ 4.2 × 10122 , which
is large enough to resist brute-force attacks [6]. For reference,
3. Key Space Analysis the key spaces of the encryption algorithms reported in [6,
Kerckhoffs’s principle is a basic principle of the modern cryp- 18, 19, 31] are 2104 , 2265 , 10117 , and 1059 , respectively, which are
tography. It illustrates that “A cryptographic system should much smaller than our key space.
be secure even if everything about the system, except the key,
is public knowledge” [45]. This implies that security of a 4. Experimental Results
cryptographic system is only dependent on secret keys, not
the algorithm itself. In other words, security of an encryption In the experiment, the parameters of our algorithm are set
algorithm mainly depends on the size of key space. In general, as follows. The initial values of Henon map are x(0) = 0.1
the larger the key space, the more secure the algorithm. The and y(0) = 0.3. The block size is S = 150 and the overlapping
key space of our algorithm includes three parts. The first sizes are t𝑥 = 70 and t 𝑦 = 70. The key of the pseudo-random
part is the initial values of the Henon chaotic map x(0) and generator is k1 = 2.
y(0), and the random block pattern determined by the block
size S and the overlapping sizes t𝑥 and t 𝑦 . As x(0) and y(0) 4.1. Encrypted Results. To validate our algorithm, some gray-
are floating numbers and require 64 bits’ storage, their key scale images and color images are selected as test images.
space is 264×2 = 2128 . The space of random block pattern is Figure 5 presents these test images and their detailed infor-
(min(𝑀, 𝑁) − 31)(𝑆 − 1)2 . The second part is the random mation is listed in Table 1. We apply our encryption scheme
key used to control the pseudo-random generator for block to these images and find that all encrypted images are chaotic
selection. The precision of the key is 64 bits. Since the key is images. Figures 6(a)–6(f) are the encrypted versions of the
Security and Communication Networks 7

Table 1: Test images.

Image Lena Fingerprint Woman Lake Goldhill ChestXray

Size 512×512 256×256 256×256 512×512 576×720 418×602
Type Grayscale Grayscale Color Color Grayscale Grayscale

Table 2: Correlation coefficients of the original images and their encrypted images.

Horizontal direction Vertical direction Diagonal direction

Test image
Original image Encrypted image Original image Encrypted image Original image Encrypted image
Lena 0.9757 −0.0685 0.9692 0.0857 0.8820 0.0059
Fingerprint 0.9142 0.0933 0.9738 0.0616 0.8050 −0.0101
Woman 0.9655 0.0334 0.9758 0.1030 0.8825 −0.0007
Lake 0.9677 −0.0062 0.9638 −0.0137 0.9144 −0.0214
Goldhill 0.9780 −0.0351 0.9625 0.0556 0.6915 0.0330
ChestXray 0.9974 0.1183 0.9969 −0.0403 0.9364 −0.0059

(a) Encrypted Lena (b) Encrypted Fingerprint (c) Encrypted Woman (d) Encrypted Lake

(e) Encrypted Goldhill (f) Encrypted ChestXray

Figure 6: Encrypted images.

six test images generated by our encryption scheme. From the 𝐸 [(𝑥 − 𝜇𝑥 ) (𝑦 − 𝜇𝑦 )]
encryption results, it is observed that our encryption scheme corr (x, y) = (8)
𝜎𝑥 𝜎𝑦
can encrypt different size images, and all the encrypted
images are noise-like images and meaningless. This means where 𝜇𝑥 and 𝜇𝑦 represent mean values of x and y, 𝜎𝑥
that our algorithm can effectively encrypt images. and 𝜎𝑦 are the standard deviations of x and y, and 𝐸[⋅]
is the expectation function. The correlation coefficient is
4.2. Correlation Analysis. The pixel correlation is the degree ranging from −1 to 1. The larger the correlation coefficient,
of association of the gray values between pixels. Generally, the stronger the correlation between two pixel sequences. For
the smaller the correlation between adjacent pixels of the a plaintext image, any two adjacent pairs of pixels usually
encrypted image is, the better the performance of the encryp- have a strong correlation. A good performance encryption
tion algorithm is. When calculating the pixel correlation in a algorithm should break such correlation.
certain direction (horizontal, vertical, or diagonal direction), In the experiments, we randomly select 3000 pairs of
several adjacent pixel pairs are randomly selected, and then adjacent pixels in horizontal, vertical, and diagonal direc-
the correlation coefficient is calculated. The formula of tions, respectively, and calculate their correlation coefficients
correlation coefficient is defined as follows. to verify the performance of our algorithm. Table 2 presents
8 Security and Communication Networks

250 250 250 250

200 200 200 200

150 150 150 150

100 100 100 100

50 50 50 50

0 0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(a) Fingerprint (b) Woman (c) Lake (d) Goldhill
250 250 250 250

200 200 200 200

150 150 150 150

100 100 100 100

50 50 50 50

0 0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(e) Encrypted Fingerprint (f) Encrypted Woman (g) Encrypted Lake (h) Encrypted Goldhill

Figure 7: Distribution of adjacent pixels in horizontal direction.

(a) Decrypted Woman (b) Decrypted Lake (c) Decrypted Goldhill (d) Decrypted ChestXray

Figure 8: Decrypted images with a wrong key: x(0) = 0.100001.

the correlation coefficients of original images and their 4.3. Key Sensitivity Analysis. A good encryption algorithm
encrypted versions. It can be found that the correlation should be sensitive to the change of secret keys. This means
coefficients of the original image are close to 1, while the that a slight difference of the keys should result in a great
correlation coefficients of the encrypted image are near 0. change in the decrypted image. If the difference between
For space limitation, some typical visual results of distri- two encrypted images is very large, it is very difficult for
bution of adjacent pixels are illustrated in Figure 7. Figures attackers to break the algorithm through differential attacks.
In the experiment, we change the initial values x(0) or y(0) of
7(a)–7(d) are the pixel pair distribution of the test images the Henon chaos system slightly and keep other decryption
(i.e., Fingerprint, Woman, Lake, and Goldhill) in horizontal keys unchanged. We use these wrong keys to decrypt Figures
direction, and Figures 7(e)–7(h) are the pixel pair distribution 6(c)–6(f) and obtain the decrypted images as shown in
of their encrypted versions in the horizontal direction. The Figure 8 and Figure 9, where Figure 8 uses a wrong key x(0)
comparison shows that the pixel pairs of original images are = 0.100001 (10−6 added) and Figure 9 uses a wrong key y(0) =
concentrated around the diagonal with 45 degrees, showing 0.300001 (10−6 added). Obviously, the decrypted results with
high correlation coefficients, while the pixel pairs of the a wrong key are still noise-like images. This indicates that our
encrypted image are uniformly distributed over the entire algorithm is key-sensitive.
value interval. This illustrates that our encryption scheme
can effectively break the correlation between adjacent pix- 4.4. Histogram Analysis. Histogram is an important statis-
els. tical feature used to show distribution of pixel values. It is
Security and Communication Networks 9

(a) Decrypted Woman (b) Decrypted Lake (c) Decrypted Goldhill (d) Decrypted ChestXray

Figure 9: Decrypted images with a wrong key: y(0) = 0.300001.

3000 7000 5000

3000 6000 4500
2500 2500 4000
5000 3500
2000 2000 3000
4000
1500 1500 3000 2500
2000
1000 1000 2000 1500
500 500 1000 1000
500
0 0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 300 0 50 100 150 200 250
(a) Red component of Lake (b) Green component of Lake (c) Blue component of Lake (d) Goldhill

Figure 10: Histograms of original images.

2500 2500 2500 4000

3500
2000 2000 2000 3000
1500 1500 1500 2500
2000
1000 1000 1000 1500
500 500 500 1000
500
0 0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(a) Red component of Lake (b) Green component of Lake (c) Blue component of Lake (d) Goldhill

Figure 11: Histograms of our encrypted images.

4000 4500
2500 2500 3500 4000
2000 2000 3000 3500
2500 3000
1500 1500 2000 2500
1000 1000 1500 2000
1500
1000 1000
500 500 500 500
0 0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250 300 0 50 100 150 200 250
(a) Red component of Lake (b) Green component of Lake (c) Blue component of Lake (d) Goldhill

Figure 12: Histograms of the encrypted images generated by [6].

often exploited to measure performance of image encryption histograms of the encrypted images generated by our algo-
algorithms. In general, an efficient encryption algorithm rithm, and Figure 12 shows the histograms of the encrypted
is expected to generate encrypted image with uniformly results generated by the encryption algorithm [6]. It can
distributed histogram. Figures 10(a)–10(c) are the histograms be observed that our histograms are almost uniformly dis-
of red, green, and blue components of the color image tributed and those histograms generated by the encryption
Lake (Figure 5(d)), and Figure 10(d) is the histogram of the algorithm [6] are rugged. Therefore, from the viewpoint of
grayscale image Goldhill (Figure 5(e)). Figure 11 presents the histogram, our algorithm is also secure.
10 Security and Communication Networks

Table 3: Comparison of the variance of histogram among different algorithms.

Variance of histogram of
Image Original variance the encrypted result
[6] [25] [31] Our
6 6
Lena 0.9725×10 0.1681×10 607.3 1339 1052.4
Fingerprint 1.5573×106 0.6517×106 34.96 277.51 329.11
Woman 0.1696×106 0.0793×106 369.33 260.1 245.37
Lake 0.7899×106 0.1514×106 3515.0 2146.5 1030.3
Goldhill 1.5506×106 0.9906×106 5538.7 2519.7 1737.4
ChestXray 6.1454×106 3.5074×106 1164.6 1164.1 1356.9
Average 1.8643×106 0.9248×106 2421.6 1284.5 958.58

Table 4: Entropy comparisons among different algorithms.

Image Original [6] [25] [31] Our

Lena 7.2185 7.3634 7.9974 7.9991 7.9992
Fingerprint 5.1141 6.9960 7.9963 7.9963 7.9964
Woman 6.8981 7.4099 7.9972 7.9986 7.9991
Lake 7.7610 7.9427 7.9985 7.9993 7.9997
Goldhill 7.5195 7.6788 7.9984 7.9993 7.9995
ChestXray 5.8733 7.3865 7.9991 7.9985 7.9989
Average 6.7308 7.4629 7.9978 7.9985 7.9988

To quantitatively analyze histograms, the variance of 256), the theoretical maximum of the entropy is 8. In general,
histogram [46] is exploited to analyze performance, which is a bigger entropy means a more secure encryption algorithm.
defined as follows. Table 4 lists entropy comparisons between our algorithm and
the compared encryption algorithms [6, 25, 31]. It can be
1 𝐿−1 𝐿−1 1 2
𝑉 (𝑍) = 2
∑ ∑ (𝑧𝑖 − 𝑧𝑗 ) (9) seen that the entropies of our algorithm are all close to 8 and
𝐿 𝑖=0 𝑗=0 2 are bigger than those of the compared algorithms [6, 25, 31].
Therefore, our algorithm is more secure than the compared
where 𝑍 = {𝑧1 , 𝑧2 , . . . , 𝑧𝐿-1 }, z𝑖 (0 ≤ i ≤ L − 1), z𝑖 is the total algorithms [6, 25, 31] from the viewpoint of entropy.
number of pixels with gray value equal to i, and L = 256 for
the grayscale image. In general, the smaller the histogram
4.6. Differential Attack. Differential attack is an effective
variance, the more secure the encrypted image. Table 3 is
method for analyzing security of cryptographic system.
histogram variance comparison between our algorithm and
Generally, attacker slightly changes pixels of a plaintext
other algorithms [6, 25, 31]. From the results, it is observed
image, generates a slightly altered encrypted image, and then
that our results are all smaller than those of the compared
analyzes the relationship between the generated encrypted
algorithms, except two cases. Specifically, our results of
image and the normal encrypted image. In practice, NPCR
Fingerprint and ChestXray are bigger than those of [25, 31].
(Number of Pixels Change Rate) and UACI (Unified Average
However, for the average variance of histogram, our result
Changing Intensity) [48, 49] are often used to evaluate the
is much smaller than those of the compared algorithms.
capability of resisting differential attack. UACI and NPCR are
It means that our algorithm has better performance than
defined as follows.
the compared algorithms [6, 25, 31] in terms of variance of
histogram. UACIR,G,B
󵄨󵄨 󸀠 󵄨󵄨
4.5. Entropy Analysis. Entropy [47] is often used to describe 1 [ 󵄨󵄨󵄨𝐶R,G,B (𝑖, 𝑗) − 𝐶 R,G,B (𝑖, 𝑗)󵄨󵄨󵄨 ]
= ∑ (11)
the uncertainty or randomness of an image. It is a useful 𝑊 × 𝐻 𝑖,𝑗 255
metric for measuring security of image encryption. It is [ ]
defined as follows. × 100%
𝐿−1
𝐻 (E) = − ∑ 𝑃 (𝑒𝑖 ) log2 𝑃 (𝑒𝑖 ) (10) ∑𝑖,𝑗 𝐷R,G,B (𝑖, 𝑗)
NPCRR,G,B = × 100% (12)
𝑖=0 𝑊×𝐻
where E = {𝑒0 , 𝑒1 , . . . , 𝑒L−1 } and P(e𝑖 ) is the possibility of the where W and H are the image width and height and
occurrence of e𝑖 . For an image with 256 gray-level (e.g., L = 𝐶R,G,B (𝑖, 𝑗) and 𝐶󸀠 R,G,B (𝑖, 𝑗) are the pixel values in the i-th
Security and Communication Networks 11

Table 5: Comparison results of NPCR and UACI (unit: %).

(1, 1) (128, 128) (360, 360) (512, 512)

Algorithm Image
NPCR/UACI NPCR/UACI NPCR/UACI NPCR/UACI
Lake 99.61/33.42 99.60/33.37 99.61/33.37 99.61/33.37
Our Goldhill 99.61/33.75 99.61/33.48 99.62/33.75 99.62/33.71
Lena 99.63/33.47 99.60/33.39 99.62/33.39 99.60/33.39
Lake 99.52/34.40 99.51/34.41 99.51/34.41 99.51/34.41
[6] Goldhill 99.56/34.31 99.59/34.36 99.59/34.36 99.56/34.36
Lena 99.54/34.67 99.55/34.64 99.54/34.68 99.54/34.66
Lake 99.61/33.48 99.60/33.53 99.59/33.55 99.61/33.48
[31] Goldhill 99.60/33.53 99.58/33.52 99.60/33.55 99.61/33.49
Lena 99.61/33.51 99.60/33.48 99.60/33.61 99.58/33.58

(a) Encryption result of the modified Lena (b) Difference between (a) and the original
encrypted Lena

Figure 13: Visual result of differential attack.

row and j-th column of two encrypted images, respectively. The encryption result is shown in Figure 13(a). The difference
If 𝐶R,G,B (𝑖, 𝑗) ≠ 𝐶󸀠 R,G,B (𝑖, 𝑗), then 𝐷R,G,B (𝑖, 𝑗) = 1. Otherwise, image between Figure 13(a) and the original encrypted result
𝐷R,G,B(𝑖, 𝑗) = 0. Note that the theoretical values of NPCR of the Lena is shown in Figure 13(b). From the result, it is
and UACI are 100% and 33.33%, respectively. Generally, the observed that even if input image is slightly changed, our
closer to the theoretical values the calculated results, the more encryption result will be greatly changed. This indicates that
secure the encryption algorithm. our algorithm is highly sensitive to pixel change. Therefore,
In the experiments, the standard images Lake, Goldhill, our algorithm can resist differential attack.
and Lena are selected as the test images, where their red According to Kerckhoffs’s principle [45], a successful
components are used. Four locations (i.e., (1, 1), (128, 128), cryptanalysis should accurately estimate secret keys (equiv-
(360, 360), and (512, 512)) are selected in the test images for alent to recovering plaintext). For chosen plaintext attack,
changing pixel values. For each test image, we just alter one attacker can choose some specific plaintexts to calculate their
pixel to generate a modified original image and then conduct corresponding ciphertexts. In practice, differential attack
encryption. Table 5 presents NPCR and UACI comparison analysis is the most common way to achieve the chosen
results among our algorithm and the compared algorithms plaintext attack [18]. As well-known, diffusion technique
[6, 31]. It can be found that all NPCR values of our algorithm can ensure security of a cryptographic algorithm against
are greater than or equal to 99.60%, which is a little greater the chosen-plaintext attack [18]. To resist this attack, in this
than or equal to those of the compared algorithms [6, 31]. For paper, we achieve diffusion by using content-based keys
UACI, all our values are close to the theoretical value 33.33%, to control the Lü chaotic map for changing pixel values.
which is almost equal to the results of [31] but a little better In addition, we exploit double spiral scans to randomly
than those of [6]. scramble pixel positions. These techniques ensure that it is
For space limitation, we only present a typical visual difficult to observe useful trace between secret keys and
result. In the experiment, we change the pixel of Lena in the plaintext/ciphertext. This means that correct key estimation is
coordinate (1, 1) (its value is changed from 169 to 170) to almost impossible in practice. Therefore, the chosen-plaintext
generate the modified Lena and encrypt it with the same key. attack is impractical for our algorithm.
12 Security and Communication Networks

(a) Density is 0.01 (b) Density is 0.05 (c) Density is 0.10

Figure 14: Decrypted images under the attack of salt and pepper noise with different densities.

(a) 40 × 40 block (b) 100 × 140 block (c) 128 × 512 block

Figure 15: Encrypted images with block missing.

(a) 40 × 40 block missing (b) 100 × 140 block missing (c) 128 × 512 block missing

Figure 16: Decrypted images with block missing.

4.7. Robustness Test. To evaluate robustness performance Figure 14. It can be observed that image noises are randomly
of our algorithm, we attack many encrypted images with distributed in the decrypted images and image qualities of
operations of salt & pepper noise and block missing, respec- the decrypted images gradually decrease with the increase
tively. It is found that our algorithm can efficiently restore of noise density. Secondly, we remove image blocks with
the decrypted images from the attacked encrypted images. different sizes from the encrypted version of Lena to generate
For space limitation, typical examples are presented here. the attacked encrypted versions as shown in Figure 15. We
Firstly, we add salt & pepper noise with different densities decrypt these attacked encrypted images and obtain the
(i.e., 0.01, 0.05, and 0.10) to the encrypted version of Lena recovered results as shown in Figure 16. It is found that
(as shown in Figure 6(a)), decrypt the attacked encrypted when the size of the missing block is small (e.g., 40 ×
images, and then obtain the decrypted images as shown in 40), visual quality of the recovered image is good and the
Security and Communication Networks 13

Table 6: Computational time comparison among different algorithms (unit: second).

Algorithm Lena Fingerprint Woman Lake Goldhill ChestXray

[6] 4.601 2.350 4.065 4.059 5.747 4.444
[25] 7.788 2.984 5.838 13.098 12.231 7.782
[31] 4.821 2.718 5.774 6.514 4.382 5.219
Our 2.792 1.755 3.264 3.860 3.759 2.100

recovered image is almost the same as its original image. As Guangxi “Bagui Scholar” Teams for Innovation and
the block size becomes large (e.g., 128 × 512), visual quality Research, the Guangxi Natural Science Foundation
of the decrypted image decreases. But the appearance of the (2017GXNSFAA198222), the Project of Guangxi Science and
original image can be easily recognized from the decrypted Technology (GuiKeAD17195062, GuiKeAD16380008), the
image. From the above results, it can be concluded that our Guangxi 1000-Plan of Training Middle-Aged/Young Teachers
algorithm is robust against salt & pepper noise attack and in Higher Education Institutions, the Guangxi Collaborative
block missing. Innovation Center of Multi-Source Information Integration
and Intelligent Processing, and the Project of the Guangxi
4.8. Computational Time Evaluation. To compare computa- Key Lab of Multi-Source Information Mining & Security
tional time, we exploit the assessed algorithms to encrypt (16-A-02-02).
the six test images, i.e., Lena, Fingerprint, Woman, Lake,
Goldhill, and ChestXray, and record the running time of each References
algorithm. All algorithms are implemented with MATLAB
R2014a and run on a computer with 3.4 GHz Intel Core [1] Z. Tang, X. Zhang, X. Li, and S. Zhang, “Robust image
i5-3570 CPU and 4.0 GB RAM. Table 6 presents compu- hashing with ring partition and invariant vector distance,” IEEE
tational time comparison among different algorithms. It is Transactions on Information Forensics and Security, vol. 11, no.
1, pp. 200–214, 2016.
observed that our algorithm runs faster than the compared
[2] Z. Tang, X. Zhang, and S. Zhang, “Robust perceptual image
algorithms [6, 25, 31]. The fast speed of our algorithm is
hashing based on ring partition and NMF,” IEEE Transactions
mainly contributed by the low complexity of double spiral on Knowledge and Data Engineering, vol. 26, no. 3, pp. 711–724,
scans. 2014.
[3] X.-Y. Wang, Y.-Q. Zhang, and X.-M. Bao, “A novel chaotic image
5. Conclusions encryption scheme using DNA sequence operations,” Optics
and Lasers in Engineering, vol. 73, pp. 53–61, 2015.
In this paper, we have proposed an image encryption algo- [4] Z. Tang and X. Zhang, “Secure image encryption without
rithm based on double spiral scans and chaotic maps. A key size limitation using Arnold transform and random strategies,”
contribution is the double spiral scans, which can efficiently Journal of Multimedia, vol. 6, no. 2, pp. 202–206, 2011.
scramble pixels of image block. Moreover, content-based keys [5] Z. Tang, X. Zhang, and W. Lan, “Efficient image encryption
are generated and used to control the Lü chaotic system, so with block shuffling and chaotic map,” Multimedia Tools and
as to ensure our sensitivity to the change of input image. Applications, vol. 74, no. 15, pp. 5429–5448, 2015.
Many experiments have been done and the results have [6] G. Zhang and Q. Liu, “A novel image encryption method based
illustrated that our algorithm has good encryption perfor- on total shuffling scheme,” Optics Communications, vol. 284, no.
12, pp. 2775–2780, 2011.
mance and outperforms some popular image encryption
[7] M. Li, H. Fan, H. Ren, D. Lu, D. I. Xiao, and Y. Li, “Mean-
algorithms.
ingful Image Encryption Based on Reversible Data Hiding in
Compressive Sensing Domain,” Security and Communication
Data Availability Networks, vol. 2018, Article ID 9803519, 12 pages, 2018.
[8] Y. Wang, Y. Zhao, Q. Zhou, and Z. Lin, “Image encryption using
The images used to support the findings of this study are partitioned cellular automata,” Neurocomputing, vol. 275, pp.
included within the article. 1318–1332, 2018.
[9] X. Y. Wang, X. Zhu, X. Wu, and Y. Zhang, “Image encryption
algorithm based on multiple mixed hash functions and cyclic
Conflicts of Interest shift,” Optics and Lasers in Engineering, vol. 107, pp. 370–379,
2018.
The authors declare that there are no conflicts of interest
[10] U. Hayat and N. A. Azam, “A novel image encryption scheme
regarding the publication of this paper.
based on an elliptic curve,” Signal Processing, vol. 155, pp. 391–
402, 2019.
Acknowledgments [11] U.S. National Institute of Standards and Technology (NIST),
“Announcing the advanced encryption standard (AES),” vol. 29,
This work is partially supported by the National Natural no. 8, pp. 2200–2203, Federal Information Processing Standards
Science Foundation of China (61562007, 61762017), Publication, 2001.
14 Security and Communication Networks

[12] S. K. Banerjee, “High speed implementation of DES,” Computers [30] R. Parvaz and M. Zarebnia, “A combination chaotic system
& Security, vol. 1, no. 3, pp. 261–267, 1982. and application in color image encryption,” Optics & Laser
[13] J. Fridrich, “Symmetric ciphers based on two-dimensional Technology, vol. 101, pp. 30–41, 2018.
chaotic maps,” International Journal of Bifurcation and Chaos, [31] H. Liu, A. Kadir, and X. Sun, “Chaos-based fast colour image
vol. 8, no. 6, pp. 1259–1284, 1998. encryption scheme with true random number keys from envi-
[14] F. Sun, S. Liu, Z. Li, and Z. Lü, “A novel image encryption ronmental noise,” IET Image Processing, vol. 11, no. 5, pp. 324–
scheme based on spatial chaos map,” Chaos, Solitons & Fractals, 332, 2017.
vol. 38, no. 3, pp. 631–640, 2008. [32] X. Chen and C.-J. Hu, “Adaptive medical image encryption
[15] Z. Liu, Q. Guo, L. Xu, M. A. Ahmad, and S. Liu, “Double algorithm based on multiple chaotic mapping,” Saudi Journal
image encryption by using iterative random binary encoding in of Biological Sciences, vol. 24, no. 8, pp. 1821–1827, 2017.
gyrator domains,” Optics Express, vol. 18, no. 11, pp. 12033–12043, [33] X. Chai, X. Zheng, Z. Gan, D. Han, and Y. Chen, “An image
2010. encryption algorithm based on chaotic system and compressive
[16] G. Chen, Y. Mao, and C. K. Chui, “A symmetric image encryp- sensing,” Signal Processing, vol. 148, pp. 124–144, 2018.
tion scheme based on 3D chaotic cat maps,” Chaos, Solitons & [34] J. Wu, X. Liao, and B. Yang, “Image encryption using 2D Hénon-
Fractals, vol. 21, no. 3, pp. 749–761, 2004. Sine map and DNA approach,” Signal Processing, vol. 153, pp.
[17] J. Wu, X. Liao, and B. Yang, “Cryptanalysis and enhancements 11–23, 2018.
of image encryption based on three-dimensional bit matrix [35] P. Singh, A. K. Yadav, and K. Singh, “Phase image encryption
permutation,” Signal Processing, vol. 142, pp. 292–300, 2018. in the fractional Hartley domain using Arnold transform and
[18] C. Fu, G. Zhang, M. Zhu, Z. Chen, and W. Lei, “A New singular value decomposition,” Optics and Lasers in Engineering,
Chaos-Based Color Image Encryption Scheme with an Effi- vol. 91, pp. 187–195, 2017.
cient Substitution Keystream Generation Strategy,” Security and [36] S. Vashisth, H. Singh, A. K. Yadav, and K. Singh, “Image
Communication Networks, vol. 2018, Article ID 2708532, 13 encryption using fractional Mellin transform, structured phase
pages, 2018. filters, and phase retrieval,” Optik - International Journal for
[19] J. Wu, X. Liao, and B. Yang, “Color image encryption based Light and Electron Optics, vol. 125, no. 18, pp. 5309–5315, 2014.
on chaotic systems and elliptic curve ElGamal scheme,” Signal [37] E. A. Naeem, M. M. Abd Elnaby, N. F. Soliman et al., “Efficient
Processing, vol. 141, pp. 109–124, 2017. implementation of chaotic image encryption in transform
[20] M. Amin, O. S. Faragallah, and A. A. Abd El-Latif, “A chaotic domains,” The Journal of Systems and Software, vol. 97, pp. 118–
block cipher algorithm for image cryptosystems,” Communica- 127, 2014.
tions in Nonlinear Science and Numerical Simulation, vol. 15, no.
[38] A. Belazi, A. A. Abd El-Latif, A.-V. Diaconu, R. Rhouma, and S.
11, pp. 3484–3497, 2010.
Belghith, “Chaos-based partial image encryption scheme based
[21] A. A. Abd El-Latif, X. Niu, and M. Amin, “A new image cipher on linear fractional and lifting wavelet transforms,” Optics and
in time and frequency domains,” Optics Communications, vol. Lasers in Engineering, vol. 88, pp. 37–50, 2017.
285, no. 21-22, pp. 4241–4251, 2012.
[39] M. H. Annaby, M. A. Rushdi, and E. A. Nehary, “Color image
[22] A. A. Abd El-Latif and X. Niu, “A hybrid chaotic system and encryption using random transforms, phase retrieval, chaotic
cyclic elliptic curve for image encryption,” AEÜ - International maps, and diffusion,” Optics and Lasers in Engineering, vol. 103,
Journal of Electronics and Communications, vol. 67, no. 2, pp. pp. 9–23, 2018.
136–143, 2013.
[40] M. Zhang and X. Tong, “Joint image encryption and compres-
[23] Z. Tang, J. Song, X. Zhang, and R. Sun, “Multiple-image sion scheme based on IWT and SPIHT,” Optics and Lasers in
encryption with bit-plane decomposition and chaotic maps,” Engineering, vol. 90, pp. 254–274, 2017.
Optics and Lasers in Engineering, vol. 80, pp. 1–11, 2016.
[41] C. Li, H. Li, F. Li, D. Wei, X. Yang, and J. Zhang, “Multiple-image
[24] L. Wang, H. Song, and P. Liu, “A novel hybrid color image
encryption by using robust chaotic map in wavelet transform
encryption algorithm using two complex chaotic systems,”
domain,” Optik - International Journal for Light and Electron
Optics and Lasers in Engineering, vol. 77, pp. 118–125, 2016.
Optics, vol. 171, pp. 277–286, 2018.
[25] Y. Abanda and A. Tiedeu, “Image encryption by chaos mixing,”
[42] C. Wu, Y. Wang, Y. Chen, J. Wang, and Q. Wang, “Asymmetric
IET Image Processing, vol. 10, no. 10, pp. 742–750, 2016.
encryption of multiple-image based on compressed sensing
[26] A. Belazi, A. A. Abd El-Latif, R. Rhouma, and S. Belghith, and phase-truncation in cylindrical diffraction domain,” Optics
“Selective image encryption scheme based on DWT, AES Communications, vol. 431, pp. 203–209, 2019.
S-box and chaotic permutation,” in Proceedings of the 11th
International Wireless Communications and Mobile Computing [43] J. Lü and G. Chen, “A new chaotic attractor coined,” Interna-
Conference, IWCMC 2015, pp. 606–610, Croatia, August 2015. tional Journal of Bifurcation and Chaos, vol. 12, no. 3, pp. 659–
661, 2002.
[27] A. Belazi, A. A. Abd El-Latif, and S. Belghith, “A novel image
encryption scheme based on substitution-permutation network [44] E. N. Lorenz, “Deterministic Nonperiodic Flow,” Journal of the
and chaos,” Signal Processing, vol. 128, pp. 155–170, 2016. Atmospheric Science, vol. 20, pp. 130–141, 1963.
[28] Z. Tang, F. Wang, and X. Zhang, “Image encryption based on [45] “Kerckhoffs’s principle,” 2015, http://crypto-it.net/eng/theory/
random projection partition and chaotic system,” Multimedia kerckhoffs.html.
Tools and Applications, vol. 76, no. 6, pp. 8257–8283, 2017. [46] Y.-Q. Zhang and X.-Y. Wang, “A symmetric image encryption
[29] L. Li, B. Abd-El-Atty, A. A. A. El-Latif, and A. Ghoneim, algorithm based on mixed linear-nonlinear coupled map lat-
“Quantum color image encryption based on multiple discrete tice,” Information Sciences, vol. 273, pp. 329–351, 2014.
chaotic systems,” in Proceedings of the 2017 Federated Conference [47] Z. Tang, X. Zhang, L. Huang, and Y. Dai, “Robust image hashing
on Computer Science and Information Systems, FedCSIS 2017, pp. using ring-based entropies,” Signal Processing, vol. 93, no. 7, pp.
555–559, Czech Republic, September 2017. 2061–2069, 2013.
Security and Communication Networks 15

[48] Q. Wang, Q. Guo, and J. Zhou, “Double image encryption

based on linear blend operation and random phase encoding
in fractional Fourier domain,” Optics Communications, vol. 285,
no. 21-22, pp. 4317–4323, 2012.
[49] Y. Wang, C. Quan, and C. J. Tay, “Nonlinear multiple-image
encryption based on mixture retrieval algorithm in Fresnel
domain,” Optics Communications, vol. 330, pp. 91–98, 2014.
 International Journal of

Rotating Advances in
Machinery Multimedia

The Scientific
Engineering
Journal of
Journal of

Hindawi
World Journal
Hindawi Publishing Corporation Hindawi
Sensors
Hindawi Hindawi
www.hindawi.com Volume 2018 http://www.hindawi.com
www.hindawi.com Volume 2018
2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Journal of

Control Science
and Engineering

Advances in
Civil Engineering
Hindawi Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Submit your manuscripts at

www.hindawi.com

Journal of
Journal of Electrical and Computer
Robotics
Hindawi
Engineering
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

VLSI Design
Advances in
OptoElectronics
International Journal of

International Journal of
Modelling &
Simulation
Aerospace
Hindawi Volume 2018
Navigation and
Observation
Hindawi
www.hindawi.com Volume 2018
in Engineering
Hindawi
www.hindawi.com Volume 2018
Engineering
Hindawi
www.hindawi.com Volume 2018
Hindawi
www.hindawi.com www.hindawi.com Volume 2018

International Journal of
International Journal of Antennas and Active and Passive Advances in
Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration
Hindawi Hindawi Hindawi Hindawi Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018