Accepted Manuscript

An efficient and secure chaotic cipher algorithm for image content

preservation

Souyah Amina , Faraoun Kamel Mohamed

PII: S1007-5704(17)30439-2
DOI: 10.1016/j.cnsns.2017.12.017
Reference: CNSNS 4408

To appear in: Communications in Nonlinear Science and Numerical Simulation

Received date: 12 November 2016

Revised date: 27 October 2017
Accepted date: 30 December 2017

Please cite this article as: Souyah Amina , Faraoun Kamel Mohamed , An efficient and secure chaotic
cipher algorithm for image content preservation, Communications in Nonlinear Science and Numerical
Simulation (2017), doi: 10.1016/j.cnsns.2017.12.017

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

Highlights:

 A new chaotic cipher algorithm for efficient and secure image content preservation is suggested,
the proposal is well-suited for both standard and medical images, under just one round.
 The proposal consists of two modules, which are iteratively performed: chaotic confusion and pixel
diffusion, the first module is governed by means of a nonlinear bit-level shuffling and circular-
shifting, while the second module is ruled by means of an improved expanded XOR (eXOR)
operation, the whole cryptosystem is controlled by the generated chaotic sequences of Logistic
Tent System (LTS).
 The obtained results indicate the high performance in terms of execution-time and security level of
the proposed cipher algorithm, and hence confirm its efficiency for real-time secure image
transmission.

T
IP
CR
US
AN
M
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT

An efficient and secure chaotic cipher algorithm for image content preservation
Souyah Amina1, Faraoun Kamel Mohamed2
Computer Science Department EEDIS Laboratory Djilalli Liabes University Sidi Bel Abbès, Algeria
1
[email protected] , [email protected]
Abstract.

This paper proposes a new chaotic cipher algorithm for efficient and secure image content preservation,
this method is specialized for both standard and medical images, and it consists of two modules which are
iteratively performed: chaotic confusion and pixel diffusion. An improved 1D chaotic system (i.e., Logistic
Tent System (LTS)) is employed in both confusion and diffusion modules, where its initial conditions are
dynamically generated and controlled by the external secret key and SHA-256 hash value of the plain

T
image, conducted to random-like generating key-streams, elevated the sensitivity to small changes on the

IP
plain image, and hence ensured the immunity of the proposal against known/chosen plain image attacks.
The confusion module is governed by a novel nonlinear bit-shuffling and circular-shifting technique, aiming

CR
to achieve bit balancing effect, mixing effect of the pixel value, and certain diffusion mechanism. The
diffusion module is ruled by means of an improved XOR operation (eXOR), to further promote the
sensitivity to plain image, and accelerate the diffusion mechanism of the overall cipher algorithm. Given

US
that the diffusion mechanism with respect to pixel value mixing are contributed by the two modules, only
one encryption round is needed to make a good combination between computational performance, and
sufficient security. The obtained results indicate the high performance in terms of execution-time and
AN
security level of the proposed cipher algorithm, and validate its robustness against cryptographic attacks,
and hence confirm its efficiency for real-time secure image transmission.

Keywords.
M

Medical image, chaos, an improved 1D chaotic system (LTS), nonlinear bit-level shuffling, circular-shifting
technique.
ED

1. Introduction

Since there has been a remarkable growth in computational power, and with the unstoppable and daily
PT

usage of digital images over open networks, a serious issue for researchers is opened to investigate better
solutions for secure storing and transmitting the content of these digital visual data. In particular those
involving sensitive and confidential information such as patients’ digital medical images *14], that should be
CE

only accessible by their authorized staffs. Therefore, whenever a digital image is needed to be employed
within any application, its content preservation turns to be a critical concept. Due to the specific
characteristics of digital images like large data size, bulk data capacity, low entropy, strong pixel correlation
AC

and high redundancy, traditional encryption methods such as DES, IDEA, AES, are not adequate to deal with
visual data content requirements, since they are originally designed for securing textual data [15,16].
Moreover, these traditional means seem to not be efficient especially for practical use regarding the non
reasonable speed and power for performing image encryption. To this end, numerous cryptographic
approaches for digital images have been suggested in the scientific literature relied on diverse principles
[17], among which are those incorporating the use of chaotic systems [1-2, 4-11, 14-20, 30, 33]. It is
required for a good cipher to accomplish some basic cryptographic significance, namely, confusion,
diffusion and randomness, these desirable properties can be reached with the specific features of chaotic
systems like ergodicity, sensitivity to initial conditions/ system parameters and random behavior, so that,
this type of dynamical systems seems to be a good candidate to be investigated as a base to design new
proposals, especially for practical use, as regards to their reasonable computational time, high security level
 ACCEPTED MANUSCRIPT

and complexity[19]. Most of chaotic proposals to cipher digital images typically follow the classical
architecture suggested by Fridrich [18], in which it consists of the iterative application of two phases:
confusion phase and diffusion phase. In the confusion phase, plain image’s pixels are shuffled without any
change in their values, aiming to conceal the correlations among adjacent pixels, this phase is ruled by
means of any appropriate two-dimensional chaotic system. In the diffusion phase, pixel values are modified
sequentially by means of a selected one-dimensional chaotic system, aiming to change the statistical
features of the plain image, so that, a minor alteration on this plain image leads to significant changes
appeared over the entire corresponding cipher image. In the last two decades, remarkable efforts were
devoted to improve the typical Fridrich’s paradigm, which can be viewed in various aspects: regarding to
the basic operating unit considered in the confusion phase, these proposals are divided into two
approaches; the first one is those proposals, in which the pixel is regarded as the basic operation unit [8,

T
12, 13, 17] within this strategy the pixel locations are exchanged, so that, only the information at these
locations is modified. In the second approach the bit is considered as the basic operation unit [4, 5, 7, 11,

IP
14, 15, 29], within this strategy the pixels’ bits locations are exchanged, so that, not only the pixel’s position
is modified but its value as well, aiming to overcome the limitations of pixel level permutation and

CR
introducing certain diffusion effects within just this stage. As it was pointed out by researchers in their
works, a considerable amount of computation is consumed within the diffusion phase, so aiming to
enhance this phase, in the way of reducing the number of its rounds while preserving the required security

US
level, was the subject of diverse proposals [21-24], in [21] a simple table look up and swapping techniques
are suggested to be employed in the diffusion phase, rather than the commonly use of 1D chaotic map,
with its limitations of time-consuming, regarding the real number arithmetic operation and the subsequent
AN
quantization step. A new bidirectional diffusion technique is proposed in [22-23], this latter consists of the
iterative application of two steps, within the first step, the conventional diffusion process is handled,
whereas pixel values are sequentially modified from top to bottom, right to left backward directions in the
M

second step, aiming to spread out the diffusion effect within the whole image, and hence fewer number of
rounds are needed, with lower computational time and sufficient security level. In [24], a continuous
diffusion technique is proposed, in which it consists of the iterative application of two steps, in the first
ED

step, the conventional diffusion process is performed, whereas in the second step a complementary
diffusion is added, aiming to achieve sufficient security level under few number of rounds, and hence the
time-consuming is significantly reduced. Both confusion and diffusion modules are incorporated within just
PT

one phase, instead of considering them as two separate phases, aiming to enhance the computational
time, since image scanning for obtaining pixel values is performed for only one time [25, 26]. The
incorporation of plain image’s features is introduced in the mechanism of chaos initial values computation,
CE

for the sake of generating one-time key-streams, and hence ensuring the resistance of the cryptosystem to
known/chosen plain image attacks [2, 27, 28]. Combined cellular automata and chaos in the design of new
proposals is suggested in [12-13, 27], and improved key-stream generators [15, 29].
AC

On the other side, the visual characteristics of medical images are completely different from the usual
standard images, in which a medical image contains more than 70% of 0’s bits [6]. In this context, those
proposals with promoted focus to higher bit-planes, and less focus to lower bit-planes [5], in term of
security, are poorly suited for this particular kind of images. To this end, a variety of chaos-based medical
image encryption are proposed [6, 14, 20]. In [14], the suggested scheme consists of two iterative phases,
namely, permutation phase and substitution phase. The former is based on a bit-level shuffling mechanism,
and governed by Arnold cat map, in which a certain substitution module is introduced within this phase. As
a latter part, substitution phase is performed by means of snake-like mode and logistic map. The suggested
cipher algorithm is iterated for certain number of rounds, to meet sufficient security level. In [20], the
suggested scheme is composed of two phases, namely, shuffling phase and masking phase, the encryption
 ACCEPTED MANUSCRIPT

is block-based and it is performed by means of chaotic cat maps, a mixing technique is added as a final step
to ensure the resistance to the commonly cryptanalysis attacks like differential ones, both full and selective
encryption can be handled by the suggested scheme, where several rounds are needed to achieve
sufficient security level. In [6], the suggested scheme consists of two iterative modules, namely,
permutation module and diffusion module. The former is ruled by nonlinear inter-pixel computing and
swapping procedure (NICSP) as a new confusion technique, in which certain diffusion effect can be
introduced within this phase. As a later part, diffusion phase is governed by snake-like mode, realized in the
reverse order. The two phases are based on chaotic logistic map, where two rounds NICSP and one
diffusion module between them are needed, to withstand known/chosen plain image attacks and meet a
sufficient security level. To meet these challenges, there is an urgent need for an adequate and efficient
medical image encryption scheme, which considers the special features of medical images, and responds to

T
the ever-growing demand for real-time teleradiology and the other telehealth applications. To this end, a
new approach is proposed within this paper for efficient and secure image content preservation, this

IP
proposal is well-suited for both standard and medical images, and it is comprised of the iterative
application of two modules: chaotic confusion and pixel diffusion. An improved 1D chaotic system (i.e.,

CR
Logistic Tent System (LTS)) proposed in [1] is employed in both confusion and diffusion modules, the
chaotic behavior of this latter is studied in detail in [1], where excellent chaotic properties are obtained,
and the well-known limitations of 1D chaotic systems are overcome, hence, the computational

US
performance of the proposed approach is enhanced by evading the use of higher-dimensional chaotic
systems. The dynamical initial conditions of the employed 1D chaotic system are produced by means of the
external secret key of 256 bit-length, and the 256-bit long hash value of the plain image, such dependency
AN
conducted to random-like generating key-streams, promoted the sensitivity to minor changes of the plain
image, and guaranteed the resistance to known/chosen plain image attacks (CPA secure). In the confusion
phase, a novel nonlinear bit-level shuffling and circular-shifting strategy is developed, in which it consists of
M

two iterative sub phases: in the former sub phase, each bit within the pixel is exchanged based on the sum
of its three composed neighborhood (i.e., the bit considered to be relocated, its immediate left-bit and
right-bit, and hence 23 possible positions can be appeared), aiming to introduce the mixing effect of the
ED

pixel value; in the later part sub phase, image’s rows and columns are circularly shifted, and governed by
the generated chaotic sequences, aiming to handle a certain diffusion mechanism, and achieve the bit
balancing effect. The diffusion phase is governed by an improved XOR operation (eXOR), and controlled by
PT

means of the generated chaotic sequences, to further promote considerably the sensitivity to plain image,
and thus accelerate the diffusion mechanism of the overall cipher algorithm. The proposed approach
possesses a good combination between computational performance, and sufficient security, since it is
CE

performed under one encryption round (one confusion round, one diffusion round), and the diffusion
mechanism with respect to the pixel value mixing effect are contributed by the two phases.
AC

The rest of the paper is organized as follows. Section 2 presents the basic definitions concerning an
improved 1D chaotic system and expanded XOR operation. The detailed nonlinear bit-level shuffling and
circular shifting based confusion, and pixel diffusion modules are discussed in section 3. Section 4 exhibits
the effectiveness of the newly proposed confusion technique. Security analysis and Extensive performance
evaluation of the proposed cipher algorithm are analyzed in detail within section 5. Finally, section 6 gives a
conclusion to the paper.

2. The Mathematical Background

In this section, the mathematical preliminaries around an improved 1D chaotic system, and an improved
expanded XOR (eXOR) operation is introduced.
 ACCEPTED MANUSCRIPT

2.1 Chaotic behavior of the improved 1D chaotic system

In [1], three improved 1D chaotic systems, namely, Logistic-Tent system (LTS), Logistic-Sine system (LSS)
and Tent-Sine system (TSS), were suggested in the scientific literature. Each chaotic system is a nonlinear
mixture of two different and existing 1D chaotic maps, i.e., Logistic map, Tent map and Sine map, which are
considered as seed maps. Logistic-Tent system (LTS) is arbitrary chosen to be used in this paper, it is based
on the nonlinear combination of both Logistic map and Tent map as seed maps, and it is describes as
follows:
( ) ( ( ) (( ) )) (1)

( ( ) ( ) ) (2)
{
( ( ) ( )( ) )

T
IP
Where ( -, and Where ( ) and (( ) ) are two existing 1D chaotic maps (seed maps),
which are Logistic map and Tent map, respectively.

CR
Simulations and analysis in [1] have clarified the excellent chaoticness that characterize LTS, LSS and
TSS. Moreover, these improved 1D chaotic systems own at least three advantages in contrast with their

US
corresponding seed maps: (i) the chaotic sequences of the suggested chaotic systems own a uniform
distribution within [0,1], in contrast with their corresponding seed maps with a restricted data ranges
within [0,1]; (ii) these 1D improved chaotic systems possess a larger chaotic range; besides to their (iii)
AN
larger Lyapunov exponents in comparison with their corresponding seed maps, revealing the better
chaoticness behavior.

2.2 The improved expanded XOR operation

M

The improved expanded XOR operation [3, 4] (eXOR, for short) is presented for the sake of increasing
the plain image’s sensitivity, and enhancing the overall security level. For two inputs ∑ and
∑
ED

, the eXOR operator is defined as follows:

(3)
( ) ∑ ( )
PT

Where not (x) flips a single bit x, denotes the XOR operation, and denotes the multiplication operation.
The operator is featured by the giving property: if eXOR(x,r)=t, then eXOR(t,r)=x. Table 1 shows clearly the
CE

deduced property.
Table 1. The results of ( ).
AC

00 01 10 11
0 1 0 0 1
1 0 1 1 0

3. The proposed Cipher Algorithm

A new encryption algorithm for efficient and secure image content preservation is presented in figure 1.
This method is specialized for both standard and medical images, and it consists of two iterative phases:
chaotic confusion and pixel diffusion. An improved 1D chaotic system, namely, Logistic Tent System (LTS) is
employed by both of these phases, where its dynamical initial value and system parameter are produced by
means of the external secret key, and 256 bit long hash value of the plain image, for the sake of generating
 ACCEPTED MANUSCRIPT

one-time chaotic sequences, increasing the sensitivity to small changes of the plain image, and hence
ensuring the immunity of the cryptosystem against known/chosen plain image attacks. The confusion
phase is governed by means of a novel nonlinear bit-shuffling and circular shifting technique, in which it
composes of two iterative sub phases: in the first sub phase, every pixel’s bit is relocated based on the sum
of its three composed neighborhood (i.e., the bit considered to be relocated, its immediate left-bit and
right-bit, and hence 23 possible positions can be appeared), to further introduce the mixing effect of the
pixel value; in the later part sub phase, image’s rows and columns are rotated, and controlled by the
generated chaotic sequences, aiming to handle a certain diffusion mechanism, and achieve the bit
balancing effect. The diffusion phase is ruled by means of an improved XOR operation (eXOR), aiming to
elevate the sensitivity to plain image, and accelerate the diffusion mechanism of the whole cipher
algorithm. A good combination between computational performance, and sufficient security is achieved by

T
our proposal under just one encryption round. The detailed description of the encryption algorithm is given
below.

IP
CR
US
AN
M
ED
PT

Figure.1 Architecture of the proposed cipher algorithm.

CE

3.1 Generation of the variable initial value and control parameter

AC

In cryptography, SHA-256 is a commonly employed cryptographic hash function with 256-bit hash value
[2]. The proposed cryptosystem uses 256-bit as external secret key, which can be generated by means of
any cryptographic secure pseudorandom bit generator, this later together with the 256-bit hash value of
the plain image are Xored to obtain a dynamic secret key K. Hence, a cryptosystem with a total complexity
of 2256, is enormous enough to withstand any kind of brute force Attacks.
The 256-bit dynamic key is divided into 8-bit blocks ( ) such as:
(4)
The initial value and the control parameter can be extracted as follows:
( ) (5)
( )
 ACCEPTED MANUSCRIPT

( ) (6)
( )

Where and are the given initial values.

Obviously, the new initial values of LTS chaotic system, namely and are greatly sensitive to a
minor change at a bit-level applied to either the secret key or the plain image, hence, they are both secret
key/plain image related, leading to the production of dissimilar chaotic sequences for different plain images
even with identical set of keys.

3.2 Nonlinear bit-level shuffling and circular-shifting based confusion phase

T
In [5], Zhu et al. pointed out their conclusion of that, different information content is carried by bits at
different positions in a pixel. As an example, a “1” at the 8th bit’s location within a pixel represents 128 (27),

IP
however at the 1st bit’s location it only represents 1 (20). The percentage of information p(i) provided by the
ith bit is computed using Eq.(7), and listed in Table 2.

CR
Table 2. The percentage of information provided by each pixel’s bit.

Bit location
0
1
2
US
The percentage of information p(i) (%)
0.3922
0.7843
1.5685
AN
3 3.137
4 6.275
5 12.55
6 25.10
M

7 50.20
ED

() , i={0,1,…,7} (7)
∑

From Table 2, it is obvious that the higher 4 bits carry 94.125 % of the total pixel’s information, however
PT

with the lower 4 bits, this pixel’s information is less than 6%.
Based on this conclusion, a permutation method at a bit-level (BLP) is suggested [5], in which the higher
4 bits are permuted individually, whereas the remaining 4 lower bits are considered as a whole and
CE

relocated, in order to reduce the computation cost.

On the other hand, the computed information percentage of different pixel’s bits using Zhan’s equation,
can’t reflect the real information distinguished by the human eyes, more precisely in case of medical
AC

images, in which these later possess more than 70 % of 0’s bits, and by the way are quite different from the
usual standard images [6]. Moreover, the lower bit-planes of medical images should be treated similarly as
higher bit-planes in term of security level for the sake of sensitive information content preservation. In this
sense, image encryption proposals with elevated concentration to higher bit-planes, and less attention to
lower bit-planes [5] are not appropriated for this special kind of images. To address the above-mentioned
limitations, a new permutation method is proposed, in which similar level of security is provided for each
pixel’s bit constitution. Hence, the suggested bit- permutation is well-matched for both standard images
and medical images. Figure 2 and figure 3 display the set of the employed standard 256 grayscale images of
size, and medical 256 grayscale images of size, respectively.
 ACCEPTED MANUSCRIPT

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

T
IP
CR
(e) (f) (g) (h)

US
Figure.2 The standard test images: (a) Lena; (b) Peppers; (c) Baboon; (d) Lake; (e) Boats; (f) Bridge; (g)
Goldhill; (h) Barbara .
AN
M
ED

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

PT
CE
AC

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

Figure.3 The medical test images: (a) CT_Abdomen; (b) CT_Hand; (c) CT_Head; (d) CT_Paranasal_sinus; (e)
MR_Brain; (f) MR_Cervital_vertibra; (g) MR_Knee; (h) X_Chest .

The steps of the newly proposed confusion strategy are described as follows:

Input: gray-scale image P of size , in which is the number of rows, is the number of columns,
the computed initial condition and control parameter , following Sect. 3.1.
Output: the confused image P’’.
 ACCEPTED MANUSCRIPT

Step 1: Get the hash value of the gray-scale image P, together with the utilized 256-bit external secret key
are Xored to obtain the dynamic key K, this later is used to calculate the initial values of Eq.(2) through Eqs
(4) and (5).
Step 2: Iterate Eq.(2) for l (l >= 500) times to avoid the transient effect, using the updated initial condition
and control parameter . Continue to iterate the LTS chaotic system for N times, times and
times, to obtain three chaotic sequences with length , and respectively i.e.,
* +, * + and
* +
Step 3: in the bit-shuffling process, the plain image P and the shuffled image P’ with size are viewed
as two-dimensional arrays, and the generated chaotic sequence init-conf is used to relocate the bits of each

T
plain image’s pixel as follows:
1) For i=1 : M

IP
a. For j=1 : N , read pixels for each plain image’s row,
( ) (8)

CR
b. Get the equivalent binary array of , denoted by
c. Each ’s bit is exchanged based on the neighborhood’s sum of the bit considered to

US
be relocated. The neighborhood is composed of three adjacent bits extracted from init-
conf, and it is consisted of: the bit located at the same position of the bit considered to be
relocated (c-bit), its immediate left-bit (l-bit) and right-bit (r-bit), the new position (NB) is
AN
calculated as follows:
NB=(l-bit * 22 + c-bit * 21 + r-bit * 20) (9)
M

To solve the problem of boundaries, the bits of the array are concatenated
together in a cyclic form to establish periodic boundary condition. Moreover, the
reversibility of the confusion phase is always verified, although having the same NB value
ED

for different c-bit positions within .

d. Modular addition operation is performed to each P’s row as follows:
( ) (10)
PT

Where is the resulted row after applying Step 3.1.c, init-conf is the generated chaotic
sequence, mod is the module operation, and is the resulted row after applying Step
CE

3.1.d.
e. And the init-conf is updated as follows:
(11)
AC

Step 4: in the circular-shifting process, the shuffled image P’1 and the confused image P’’ with size
are viewed as two-dimensional arrays. Each shuffled image’s row and column are rotated based on the
generated chaotic sequences and respectively, and it is described as follows:
1) For i=1 : M
a. For j=1 : N , read pixels for each shuffled image’s row,
( ) (12)

b. Get the equivalent binary array of denoted by

c. Row’s circular shifts is performed as follows:
( ) (13)
 ACCEPTED MANUSCRIPT

Where the value determines the number in which circular shifts of should be
rotated through itself, k is a loop variable ranging from 1 to , yielding .
d. The direction of row’s circular shifts is computed as follows:
( ) (14)

2) For j=1 : N
a. For i=1 : M , read each column’s pixels from the resulted image of step 4.1, denoted by,P’2
( ) (15)

b. Get the equivalent binary array of denoted by

c. Column’s circular shifts is performed as follows:

T
( ) (16)
Where the value determines the number in which circular shifts of should be

IP
rotated through itself, k is a loop variable ranging from 1 to , yielding .
d. The direction of column’s circular shifts is computed as follows:

CR
( ) (17)
If then right circular shifts is performed to image’s rows and columns, with and times
respectively, else left circular shifts is applied instead.

US
If the value within / is divisible by eight, then row/column rotating procedure is viewed as pixel level
permutation i.e., pixels are circularly rotated through the ith row and jth column with /8 and / steps
respectively.
AN
If the value within / is not divisible by eight, then row/column rotating procedure is viewed as bit level
permutation i.e., bits are circularly rotated from one pixel to another without any limitations.

3.3 Diffusion phase

M

Every confused image’s pixel is subjected to be diffused by means of expanded XOR operation. Aiming
ED

to, improve the overall security level, increase the plain image sensitivity for amplifying the resistance to
known/chosen plain image attacks and accelerate the diffusion effect within only one encryption round.
The steps of the improved diffusion strategy are described as follows:
PT

Input: the confused image P’’.

Output: the cipher image C.
CE

Step1: the confused image with size is transformed to one dimensional array,
.
Step2: generating one chaotic sequence using LTS chaotic system, and extracting the other one
AC

based on .
⌊ ⌋ (18)
( ) (19)

Where i is a loop variable, ranging from 1 to , , , refers to the complement

code of , ⌊ ⌋ refers to obtain the nearest integer less than or equal , denotes the XOR operation and
RCS (a,b) denotes the b-bit Right Circular Shift operation applied on the number a.

Step3: diffuse each confused image’s pixel, in which the first pixel should be computed in a specific way as
follows:
 ACCEPTED MANUSCRIPT

. ( ) ( )/ (20)

. ( ) ( )/ (21)
Where , , and , - in which it is served as a part of the secret key.
3.4 The decryption procedure

The decryption procedure is the same as that of the encryption one described above, however it must
be performed in the reverse order. First of all, the SHA-256 hash value of the plain image should be
transferred with the external secret key to the decryption part, in order to produce the initial conditions of
1D LTS chaotic system, and hence generate the required chaotic sequences. The decryption starts by
performing the diffusion phase as the first step, by following Eqs.(22)-(25), after nonlinear bit-level shuffling

T
and circular-shifting based confusion phase as the latter part step. As the confusion phase consists of two
iterative sub phases, they also must be applied reversibly, starting with circular-shifting sub phase as the

IP
first step, image’s columns and rows are circularly shifted, where the circular shift directions are switched
(i.e., If then left circular shifts is performed to image’s columns and rows, with and times

CR
respectively, else right circular shifts is applied instead); as the later part bit-shuffling sub phase is carried
out starting by the last image’s row to the first image’s row. Due to the symmetric nature of encryption and
decryption processes, they are essentially similar in their complexity and execution-time.

( (
(
US ))
)
(22)
(23)
AN
( ( )) (24)
( ) (25)
M

4. Effectiveness of the proposed confusion phase

ED

4.1 The uniformity of bit distribution within each pixel’s bit

PT

The uniformity distribution of each bit-plane within a pixel is a needed property that should be
accomplished by the permutation technique [7]. To this end, the bit distribution of different bit-planes for
both standard and medical images is experimented. Table 3 and 4 clarify the obtained results using the two
CE

testing images, namely, the standard 256 grayscale Lena image under one permutation round,
and the 256 grayscale medical CT image of Abdomen (i.e., CT-Abdomen) under two
permutation rounds, respectively. One can conclude that the suggested permutation technique is very
AC

effective, to balance the percentage of ‘0’ for both kind of images.

Table 3. The percentage of ‘0’ of Lena plain vs. permuted standard image.

‘Lena’ 8th bit 7th bit 6th bit 5th bit 4th bit 3rd bit 2nd bit 1st bit
Plain (%) 48.8830 57.9910 49.3618 49.7234 50.1136 50.1579 49.9458 49.9114
Permuted (%) 49.9816 49.8458 49.9176 49.8592 49.9027 49.8847 49.8668 49.9191

Table 4. The percentage of ‘0’ of CT-Abdomen plain vs. permuted medical image.

‘CT-Abdomen’ 8th bit 7th bit 6th bit 5th bit 4th bit 3rd bit 2nd bit 1st bit
Plain (%) 80.3260 86.5848 86.8686 86.1663 82.7411 96.0250 72.9679 76.1497
 ACCEPTED MANUSCRIPT

Permuted (%) 50.0785 50.1136 50.1247 50.1434 50.1209 50.1022 50.1228 49.9729

4.2 Pixel correlation and diffusion effect performance

One of the major roles of the permutation technique is to reduce the high redundancies between
adjacent pixels. To evaluate such property, correlation of adjacent pixels is handled, in which: 3000 pairs of
neighboring pixels in the plain image and its corresponding cipher image are randomly selected (in
horizontal, vertical and diagonal directions). The correlation coefficient for each pair is computed by means
of the following equations:

( ) (26)

T
√ ( ) √ ( )
(27)

IP
( ) ∑( ( )) ( ( )) ( ) ∑ ( ) ∑( ( ))

CR
Where and stand for the gray-level values of the ith pair of the chosen neighboring pixels within the
image, and is the whole number of samples.
Another desirable property that should be satisfied by the permutation technique is that, not only the

US
pixel’s position should be modified but also its value [7], to further accomplish both confusion and diffusion
within just this step. To testify the diffusion effect of the proposed permutation strategy, two commonly
metrics, namely, NPCR (number of pixels change rate) and UACI (unified average changing intensity), are
AN
handled to examine the influence of a minor change (i.e., at a bit-level) applied to the plain image on the
whole cipher image. The NPCR and UACI values are computed by means of the following equations:
(28)
M

( ∑ ∑ , -)

(29)
| , - , -|
ED

( ∑∑ )

Where N and M denote the image’s width and height respectively , C1[i,j] represents the obtained cipher
PT

image from the original plain image, whereas C2[i,j] is obtained after one bit plain image’s modification , for
each (i,j) position if C1[i ,j]=C2 [i,j] , then D[i,j]=0 , else D[i,j]=1.
The two experiments have been conducted for a variety of standard and medical images with different
CE

sizes using the suggested permutation technique. Table 5 and 6 illustrate the simulation results for both the
suggested permutation, and some other existing permutation techniques in the scientific literature. For the
sake of simplicity, two testing images are selected: the standard 256 gray scale Lena image , and
AC

the medical image 256 gray scale CT-Abdomen , respectively. Figure 4 and 5 show the obtained
permuted images using one round (a), two rounds (b) and three rounds (c), for both the standard Lena
image and medical CT-Abdomen image, respectively.
Table 5. Simulations of the proposed permutation and other permutation techniques using standard Lena
image.

Permutation Rounds Pixel correlation NPCR (%) UACI (%)

technique
Horizontal Vertical Diagonal
Plain image - 0.9891 0.9847 0.9813 - -
Proposed 1 0.0111 0.0076 -0.0097 41.13 10.40
Proposed 2 -0.0206 0.0100 0.0110 99.50 33.10
 ACCEPTED MANUSCRIPT

Proposed 3 -0.0219 0.0149 -0.0025 99.60 33.46

Ref.[8] 1 -0.0032 0.0029 -0.0019 73.38 15.87
Arnold Cat map Ref[8] 3 0.0276 0.0226 0.0199 3.8147e-006 1.4960e-008

Table 6. Simulations of the proposed permutation and other permutation techniques using medical CT-
Abdomen image.

Permutation Rounds Pixel correlation NPCR (%) UACI (%)

technique
Horizontal Vertical Diagonal
Plain image - 0.8615 0.6870 0.6512 - -
Proposed 1 0.1020 0.0192 -0.0049 6.22 1.45

T
Proposed 2 0.0098 0.0134 0.0048 98.63 32.64
Proposed 3 0.0169 0.0058 0.0180 99.61 33.50

IP
BLP [5] 1 -0.0032 0.0029 -0.0019 3.8147e-006 1.4960e-008
Ref.[6] 1 -0.0134 0.0024 -0.0170 7.90 2.62

CR
Ref.[6] 2 -0.0164 0.0083 0.0080 99.22 32.72
ES [7] 1 0.0551 0.1720 0.0538 3.8147e-006 1.4960e-008

US
AN
M

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

ED

Figure.4 The application of the proposed permutation technique: (a) Lena grayscale standard test image
pixels; (b) the permuted image after one round; (c) the permuted image after two rounds; (d)
the permuted image after three rounds.
PT
CE
AC

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

Figure.5 The application of the proposed permutation technique: (a) CT_Abdomen grayscale medical test
image pixels; (b) the permuted image after one round; (c) the permuted image after two
rounds; (d) the permuted image after three rounds.

With regard to the obtained results, the effectiveness and security performance of the suggested
permutation technique can be proved in three aspects: (1) not only the pixel’s location is modified but also
 ACCEPTED MANUSCRIPT

its value as well, as a consequent, the bit distribution of the confused images exhibits a good performance
of 0-1 ratio; (2) the confused images are totally indistinguishable after only one permutation round, and the
relations among adjacent pixels are significantly de-correlated; (3) both confusion and diffusion are
achieved within just this phase, in which the influence of a minor change by unity of one bit flipping applied
to the plain image, can be extended to a larger scale on the corresponding cipher image using the proposed
permutation technique, this difference is expressed by means of NPCR and UACI values. In case of standard
images, these values are how about 41 % NPCR and 10 % UACI, whereas they are how about 6% NPCR and
1% UACI in case of medical images by using only one permutation round. These values can further be
improved by increasing the number of permutation rounds, where three rounds are quite enough to satisfy
the expected level of NPCR > 99.6 % and UACI > 33.4 % , for both standard and medical test images.

T
5. Security and performance analysis of the proposed Cipher Algorithm

IP
A good cipher algorithm is one in which, it has enough immunity to withstand all kinds of cryptographic
attacks. To demonstrate the sufficient security level of our proposal, extensive security analysis has been

CR
performed. In the following experiments, only one encryption round is required (one confusion, one
diffusion), to prove the effectiveness of the cipher algorithm for both standard and medical images.

5.1. Secret key space analysis

US
A key space of a cipher algorithm should be large enough, to possess the immunity against exhaustive
AN
key search. Owing to the remarkable growth in computational power, the key space has to be no smaller
than 2128 [9]. In the proposed cryptosystem, the keys are built of:
1. The bit length of the external secret key of 256 bits.
M

2. The given initial values of and for LTS chaotic system.

3. The 256-bit long hash value.
As for the initial values of and , 10-15 is considered as the used precision of floating point numbers, to
ED

conform with IEEE floating point standard. Furthermore, the security of SHA-256 with complexity of best
attack is given as SHA-256=2128 [2], besides to the used external secret key with total complexity of 2256. So,
the overall key space can be computed by the multiplication of the aforementioned possibilities as
PT

, hence, with this sufficiently enormous key space, all types of attacks are
rendered difficult and infeasible in practical time.
CE

5.2. Statistical analysis

5.2.1 Uniformity analysis

AC

Image histogram is one of the commonly used metrics for evaluating the robustness of the cipher
algorithm against statistical attacks. It shows the distribution of pixel values within an image, by plotting
the amount number of pixels within each gray level value. To realize such experiment, several standards
and medical images have been examined with their corresponding cipher images under different secret
keys. Each pair of them reveals the same result, and for each time the histogram of the cipher image is very
close to the uniform distribution, and significantly different from that of the matching plain image. Figure 6
shows the histograms of both the standard grayscale Lena image and the medical grayscale CT-Abdomen
image. Moreover, to demonstrate such obtained uniformity, the Chi-square is carried out, and it is
calculated as follows [11]:
 ACCEPTED MANUSCRIPT

( ) (30)
∑

Where denotes the total number of levels (in case of grayscale image, =256), are the occurrence
frequencies of each gray level (0-255), and represents the expected occurrence frequency of the uniform
distribution, and it is calculated as: (M is the number of rows, N number of columns
and P is the number of plane, for grayscale image P=1). Table 7 and table 8 give the obtained results of Chi-
square test (under the significant level 0.05), performed to the histograms of standard grayscale
images ( ), and the medical grayscale images ( ), respectively, whereas a
comparison in terms of such test is performed to the cipher histograms of a set of standard grayscale
images, and introduced in table 9, with some existing methods including ours. From the obtained results, it

T
is obvious that ( ) , reflecting the efficiency of the proposed cipher to conceal
the spatial redundancy of the plain image, and hence, no detectable clue can be found to apply statistical

IP
attacks.
Table 7. The Chi-square test results for standard images.

CR
Test image
Plain image Cipher image
Lena
Pepper
Baboon
Lake
116079.62
112212.17
156878.90
162376.58
US 240.91
276.79
235.93
266.95
AN
Boats 363069.90 236.09
Bridge 989010.34 209.58
Goldhill 124757.18 247.86
M

Barbara 108261.11 232.95

Table 8. The Chi-square test results for medical images.

ED

Test image
Plain image Cipher image
CT_Abdomen 31382757.26 263.11
PT

CT_Hand 15363981.14 235.37

CT_Head 5852617.54 266.91
CT_Paranasal_sinus 25012277.10 286.68
CE

MR_Brain 5978910.51 253.70

MR_Cervical_Vertibra 7675854.60 209.25
MR_Knee 7088705.81 257.56
AC

X_Chest 2439249.25 283.52

Table 9. Comparison of the Chi-square value between our proposed approach and the other proposals.

Encryption method Image Image size Chi-square

Our proposal Lena 240.91
Ref.[23] Lena 254.15
Our proposal Lena 277.53
Ref.[16] Lena 184
Ref.[34] Lena 263
Our proposal Peppers 276.79
Ref.[23] Peppers 264.92
 ACCEPTED MANUSCRIPT

Our proposal Peppers 277.41

Ref.[16] Peppers 270
Ref.[34] Peppers 274
Our proposal Lake 266.95
Ref.[16] Lake 261
Our proposal Baboon 235.93
Ref.[23] Baboon 278.88
Our proposal Baboon 261.60
Ref.[16] Baboon 259
Ref.[34] Baboon 266

T
IP
CR
(a)
US (b)
AN
M
ED

(c) (d)
PT
CE
AC

(e) (f)

(g) (h)
 ACCEPTED MANUSCRIPT

Figure.6 Histogram test of plain/cipher image: (a) Plain Lena grayscale standard test image
pixels, (b) its corresponding histogram, (c) Cipher Lena, (d) its corresponding histogram; (e) Plain
CT_Abdomen grayscale medical test image pixels, (f) its corresponding histogram, (g) Cipher
CT_Abdomen, (h) and its corresponding histogram.

5.2.2 Entropy analysis

Information entropy is the commonly used metric to express the unpredictability and randomness of
information [Ref10]. As for image information, the distribution of its grayscale values is computed, so that,
the closer entropy to its theoretical value, reflects the more uniform the distribution of image grayscale
values is. Let m denotes the information source, so the formula for measuring information entropy is:

T
(31)
( ) ∑ ( ) ( )

IP
CR
Where represents the entirely number of possible gray level values, and ( ) denotes the probability of
occurrence of the ith gray value. The utmost achievable value of entropy i.e., the theoretical value for a
source emitting , which is the case of grayscale images, is 8. The closer cipher image’s entropy value

US
is to 8, the more unpredictability and randomness is reflected by cipher algorithm output. Table 10 and 11
give the obtained results of entropy test carried out for the selected standard test images and medical test
images, respectively, while table 12 presents the obtained experimental values for different standard
AN
cipher images, attained by applying some existing methods with respect to our proposal, thus it is obvious
that these values are extremely nearby to the maximum value of 8, rendering the information leakage from
the proposed cipher algorithm insignificant.
M

Table 10. Information entropy test results for standard images.

Test image Plain images Cipher images

ED

Lena 7.445507 7.999338

Peppers 7.571478 7.999240
Baboon 7.357949 7.999350
Lake 7.484499 7.999267
PT

Boats 7.123758 7.999350

Bridge 5.705560 7.999424
Goldhill 7.477780 7.999317
CE

Barbara 7.466426 7.999357

Table 11. Information entropy test results for medical images.

AC

Test image Plain images Cipher images

CT_Abdomen 2.270036 7.999277
CT_Hand 2.570225 7.999352
CT_Head 5.478837 7.999263
CT_Paranasal_sinus 2.618263 7.999212
MR_Brain 3.991414 7.999299
MR_Cervical_Vertibra 3.706131 7.999424
MR_Knee 4.073966 7.999291
X_Chest 4.778157 7.999219
 ACCEPTED MANUSCRIPT

Table 12. Comparison of the entropy value between our proposed approach and the other proposals.

Encryption method Image Image size Entropy

Our proposal Lena 7.999338
Ref.[8] Lena 7.999319
Ref.[17] Lena 7.999324
Ref.[23] Lena 7.999301
Ref.[28] Lena 7.999286
Our proposal Lena 7.996951
Ref.[35] Lena 7.997

T
Ref.[36] Lena 7.9972
Ref.[37] Lena 7.9973

IP
Our proposal Peppers 7.999240
Ref.[38] Peppers 7.999275

CR
Our proposal Peppers 7.99694
Ref.[35] Peppers 7.9973
Ref.[37] Peppers 7.9975
Our proposal Barbara 7.999357
Ref.[8]
Ref.[38]
Our proposal
Barbara
Barbara
Barbara
US 7.999240
7.999211
7.997047
AN
Ref.[35] Barbara 7.9971
Our proposal Baboon 7.999350
Ref.[23] Baboon 7.999263
Ref.[33] Baboon 7.9993
M

Ref.[38] Baboon 7.999345

Our proposal Bridge 7.999424
Ref.[28] Bridge 7.999268
ED

5.2.3 Correlation analysis

PT

One of the main features of an image with its meaningful visual content, is the high correlation and
redundancy among its neighboring pixels, either in horizontal, vertical or diagonal directions. An efficient
cipher algorithm should conceal such relations between adjacent pixels, and exhibit a good performance of
CE

balanced 0-1 ratio and zero correlation [12]. Table 13 and 14 give the obtained results of correlation
coefficient values of the selected standard test images and medical test images, respectively. Moreover,
correlation in all directions, have been plotted in figure 7 for the standard grayscale Lena image and the
AC

medical grayscale CT-Abdomen image. From the obtained correlation coefficient results and the shown
figure, it is obvious that the high relations among plain images’ neighboring pixels (correlation coefficient
close to 1), effectively reduced in the corresponding cipher images’ pixels (correlation coefficient close to
0), using the proposed cipher algorithm, reflecting the efficiency of this later to conceal the spatial
redundancy within the cipher image’s pixels.

Table 13. Correlation test results for standard images.

Test image Plain Ciphered

Horizontal Vertical Diagonal Horizontal Vertical Diagonal
Lena 0.9891 0.9847 0.9813 0.0119 0.0092 0.0013
 ACCEPTED MANUSCRIPT

Pepper 0.9174 0.9741 0.9047 0.0115 0.0109 -0.0101

Baboon 0.6105 0.6597 0.5203 0.0201 -0.0326 0.0097
Lake 0.8731 0.9578 0.8582 0.0045 0.0049 0.0041
Boats 0.9709 0.9365 0.9403 0.0023 0.0186 0.0099
Bridge 0.8373 0.8202 0.7548 -0.0011 0.0033 -0.0022
Goldhill 0.9782 0.9901 0.9761 0.0111 0.0032 0.0047
Barbara 0.9497 0.9034 0.8525 0.0136 0.0084 0.0146

Table 14. Correlation test results for medical images.

Test image Plain Ciphered

Horizontal Vertical Diagonal Horizontal Vertical Diagonal

T
CT_Abdomen 0.8615 0.6870 0.6512 0.0191 0.0063 0.0062
CT_Hand 0.9967 0.9754 0.9722 0.0120 -0.0035 -0.0067

IP
CT_Head 0.9678 0.9890 0.9548 -0.0061 0.0097 0.0191
CT_Paranasal_sinus 0.4781 0.7722 0.3343 0.0084 -0.0032 -0.0108

CR
MR_Brain 0.9013 0.8519 0.8264 0.0023 0.0107 0.0028
MR_Cervical_Vertibra 0.9630 0.9831 0.9506 0.0145 -0.0045 0.0024
MR_Knee 0.9626 0.9577 0.9413 0.0033 0.0092 -0.0082
X_Chest 0.9726 0.9835 0.9650 -0.0029 0.0016 -0.0019

US
AN
M
ED
PT
CE

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

AC
 ACCEPTED MANUSCRIPT

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

Figure.7 Correlation diagrams of plain/cipher image: (a) Lena grayscale standard test image

T
pixels, (b) horizontal correlation, (c) vertical correlation, (d) diagonal correlation ; (e) CT_Abdomen
grayscale medical test image pixels, (f) horizontal correlation, (g) vertical correlation, (h)

IP
diagonal correlation.

CR
5.3. Sensitivity test

5.3.1 Robustness against differential attacks

US
As stated in section 3.2, two metrics, namely, NPCR and UACI are handled to evaluate the degree of the
plain image’s sensitivity by means of the proposed cipher algorithm. For the sake of ensuring the
AN
effectiveness and robustness against differential attacks, the following experiments are performed:
(i) The standard grayscale Lena image and the medical grayscale CT-Abdomen image and their eight
modified versions are enciphered by our proposal. These modified images are attained by only changing
M

the least significant bit (LSB) of the matching randomly chosen pixels, in both Lena and CT-Abdomen
images, and they are denoted by Lena_v1, Lena_v2,…, Lena_v8, and CT_Abdomen_v1,
CT_Abdomen_v2,..,CT_Abdomen_v8. The obtained results are given in Table 15 and 16, respectively.
ED

(ii) The same experiment described in section 3.2 is carried out for all the standard and medical images. The
obtained results are given in Table 17 and 18, respectively. Table 19 introduces the obtained results of such
experimental values, for different standard cipher images, attained under the application of certain existing
PT

methods including ours. As regards to the exhibited experimental results, our proposal is highly sensitive to
plain image bit modification, and hence render such type of attacks void.
CE

Table 15. NPCR and UACI tests results for cipher Lena standard image.

Test image NPCR(%) UACI(%)

Lena_v1 99.6177 33.4441
AC

Lena_v2 99.6452 33.6152

Lena_v3 99.6086 33.4647
Lena_v4 99.6494 33.4460
Lena_v5 99.6330 33.4927
Lena_v6 99.6051 33.4781
Lena_v7 99.6231 33.4646
Lena_v8 99.6418 33.4480
Average 99.6279 33.4816

Table 16. NPCR and UACI tests results for cipher CT_Abdomen medical image.

Test images NPCR(%) UACI(%)

 ACCEPTED MANUSCRIPT

CT_Abdomen_v1 99.6173 33.4353

CT_Abdomen_v2 99.6006 33.4351
CT_Abdomen_v3 99.6231 33.4832
CT_Abdomen_v4 99.6215 33.5294
CT_Abdomen_v5 99.6044 33.4546
CT_Abdomen_v6 99.6295 33.4908
CT_Abdomen_v7 99.6116 33.4029
CT_Abdomen_v8 99.6135 33.4805
Average 99.6151 33.4639

Table 17. NPCR and UACI tests results for cipher standard images.

Test images NPCR(%) UACI(%)

T
Lena 99.6452 33.6152

IP
Pepper 99.6315 33.5073
Baboon 99.6154 33.4354
Lake 99.6349 33.3988

CR
Boats 99.6284 33.5407
Bridge 99.6139 33.4014
Goldhill 99.5964 33.5166
Barbara 99.6368
US
Table 18. NPCR and UACI tests results for cipher medical images.
33.5152
AN
Test image NPCR(%) UACI(%)
CT_Abdomen 99.7920 33.5661
CT_Hand 99.5956 33.4046
M

CT_Head 99.6086 33.4911

CT_Paranasal_sinus 99.5903 33.4944
MR_Brain 99.6696 33.3811
MR_Cervical_Vertibra 99.6170 33.3992
ED

MR_Knee 99.6860 33.5320

X_Chest 99.5986 33.4658
PT

Table 19. Comparison of the NPCR and UACI values between our proposed approach and the other
proposals.
CE

Encryption method Image Image size NPCR UACI

Our proposal Lena 99.6452 33.6152
Ref.[8] Lena 99.62 33.43
AC

Ref.[11] Lena 99.607 33.463

Ref.[17] Lena 99.62 33.41
Ref.[23] Lena 99.6052 33.4111
Ref.[28] Lena 99.6215 33.4654
Our proposal Lena 99.5941 33.5052
Ref.[16] Lena 99.5894 33.4645
Ref.[35] Lena 99.655 33.516
Ref.[36] Lena 99.61 33.46
Ref.[37] Lena 99.61 33.53
Our proposal Peppers 99.6315 33.5073
Ref.[8] Peppers 99.60 33.54
Ref.[23] Peppers 99.6052 33.4372
Ref.[28] Peppers 99.6112 33.4612
 ACCEPTED MANUSCRIPT

Ref.[38] Peppers 99.6391 33.5128

Our proposal Peppers 99.5849 33.4641
Ref.[37] Peppers 99.63 33.58
Our proposal Barbara 99.6368 33.5152
Ref.[8] Barbara 99.61 33.46
Ref.[11] Barbara 99.607 33.463
Our proposal Baboon 99.6154 33.4354
Ref.[8] Baboon 99.61 33.51
Ref.[23] Baboon 99.3504 33.4520
Ref.[33] Baboon 99.6048 33.4554
Ref.[38] Baboon 99.6101 33.4354
Our proposal Baboon 99.6017 33.6287
Ref.[16] Baboon 99.6124 33.4891

T
Our proposal Boat 99.6284 33.5407

IP
Ref.[13] Boat 99.1025 33.1600
Ref.[39] Boat 99.6154 33.4654
Our proposal Boat 99.6139 33.4751

CR
Ref.[35] Boat 99.625 33.453

5.3.2 Plain image sensitivity

US
For the sake of revealing the degree of sensitivity of the proposed cipher’s input (i.e., plain image), and
how much of dissimilarities can be achieved by its output, after a minor change on its input, the following
AN
procedure is handled: two plain images and , which differ by one randomly selected bit (the Least
Significant Bit (LSB)), are enciphered to attain two cipher images and , respectively. The hamming
distance (in bits) between these two cipher images is computed as follows:
∑ (32)
M
ED

Where lb is the length in bits of the plain/cipher images.

The experiment is carried out for a set of random standard and medical images as shown in figure 10(a). All
of them exhibit identical results, in which the hamming distance between two cipher images and , is
PT

very close to 50%, and the resultant mean value is 50.0211%, which signifies that with a minor change by
unity of one bit flipping, more than 50% of the corresponding cipher image is changed. Hence, the
proposed cipher overcomes the plain image sensitivity attacks.
CE

5.3.3 Key sensitivity

AC

Another searched property for any proposed cipher algorithm is its sensitivity to secret key, and in order
to evaluate the degree of such sensitivity, the following experiments are performed:
(1) The sensitivity of the encryption process to its secret key is assessed by enciphering each plain image
within the selected standard and medical images, by means of two slightly different keys and to
attain two cipher images and , respectively.
(2) The sensitivity of the decryption process to its secret key is assessed by deciphering each attained cipher
image , by means of two slightly different keys and .

1) Key sensitivity analysis within the encryption process

 ACCEPTED MANUSCRIPT

The sensitivity to the secret key for Lena standard image and CT_Abdomen medical image are exhibited
in figure3 and 4, respectively, in which figure 8 (a)/figure 9 (a) are the attained cipher images using the key
( ), whereas figure 8 (b)/figure 9 (b) are the attained cipher images using the key
(  ), where  is a very small value called the perturbing value [13],
within our experiment, it is equal to 10-15 , we should notice that, it is the only perturbed value and the
other parameter values of the key are kept unmodified. Therefore, the percentage of difference between
each pair of cipher images and is computed as follows:
. ∑ ∑ ( , - , -)/ where ( ) { (33)

Where M is the number of image’s rows, and N is the number of image’s columns.

T
2) Key sensitivity analysis within the decryption process

IP
In the same sense, the experiment is extended to be performed for the decryption process. Figure 8 (c)/
figure 9 (c) give the decipher images using the correct key, whereas figure 8 (d)/figure 9 (d) give the

CR
decipher images using the perturbing key.
Table 20 and 21 give the obtained results of secret key sensitivity test carried out for the selected standard
test images and medical test images, respectively, where a small change is handled for just one key at a

change on the secret key. US

time, in the original key set , thus it is obvious that our proposal has enough sensitivity against any

On the other side, and for demonstrating the degree of sensitivity of the proposed cipher’s input (i.e.,
AN
the used external secret key), and how much of difference can be achieved by its output, after a minor
change by unity of one bit flipping (at a Least Significant Bit (LSB)) on its input, the same experiment
described in section 4.3.2 is carried out for 100 different keys, and the hamming distance for each pair of
M

corresponding cipher images is computed and shown in figure 10 (b). The obtained values are very close to
the optimal value of 50%, in which the resultant mean value is 50.0097%, reflecting the robustness of our
proposal against any minor change in its secret external key, and thus the avalanche effect is achieved.
ED
PT
CE
AC

(a) Enciphered with ( ) (b) Enciphered with (  )

(c) Deciphered with correct key (d) Deciphered with perturbed key)
 ACCEPTED MANUSCRIPT

Figure.8 Shows the key sensitivity experiment for Lena standard image.

T
(a) Enciphered with ( ) (b) Enciphered with (  )

IP
CR
US
AN
(c) Deciphered with correct key (d) Deciphered with perturbed key)

Figure.9 Shows the key sensitivity experiment for CT_Abdomen medical image.
M

Table 20. Key sensitivity test results for standard images.

ED

Test image Secret keys Difference rates (%)

(  ) Encryption Decryption
Lena 99.5937 99.5964
Pepper 99.6086 99.5994
PT

Baboon 99.6006 99.6208

Lake 99.6013 99.6196
Boats 99.6131 99.6032
CE

Bridge 99.5899 99.6086

Goldhill 99.6322 99.5937
Barbara 99.6128 99.6040
 )
AC

Lena ( 99.5967 99.6212

Pepper 99.6276 99.5903
Baboon 99.6036 99.5903
Lake 99.5800 99.6025
Boats 99.6170 99.6089
Bridge 99.6006 99.6322
Goldhill 99.6112 99.5922
Barbara 99.6177 99.6025
Lena (  ) 99.5944 99.6154
Pepper 99.6078 99.6395
Baboon 99.6253 99.6124
Lake 99.6139 99.6185
Boats 99.6070 99.5975
 ACCEPTED MANUSCRIPT

Bridge 99.6166 99.6021

Goldhill 99.6158 99.6109
Barbara 99.6097 99.6112
Lena (  ) 99.6082 99.6036
Pepper 99.6101 99.5998
Baboon 99.5906 99.6131
Lake 99.6212 99.5807
Boats 99.6192 99.6150
Bridge 99.6128 99.6002
Goldhill 99.5990 99.5971
Barbara 99.6009 99.6181

T
Table 21. Key sensitivity test results for medical images.

IP
Test image Secret keys Difference rates (%)
(  ) Encryption Decryption

CR
CT_Abdomen 99.6128 99.6059
CT_Hand 99.6189 99.6185
CT_Head 99.5956 99.6223
CT_Paranasal_sinus
MR_Brain
MR_Cervical_Vertibra
MR_Knee
US
99.6047
99.6231
99.6055
99.5971
99.6192
99.5941
99.6044
99.6089
AN
X_Chest 99.6372 99.6204
CT_Abdomen (  ) 99.6009 99.5834
CT_Hand 99.5975 99.5990
M

CT_Head 99.6170 99.5941

CT_Paranasal_sinus 99.6059 99.6334
MR_Brain 99.6055 99.6318
ED

MR_Cervical_Vertibra 99.6021 99.6204

MR_Knee 99.6173 99.6032
X_Chest 99.6322 99.6170
CT_Abdomen (  ) 99.6196 99.5910
PT

CT_Hand 99.6238 99.6009

CT_Head 99.6139 99.6109
CT_Paranasal_sinus 99.5971 99.6131
CE

MR_Brain 99.5994 99.6002

MR_Cervical_Vertibra 99.6105 99.6250
MR_Knee 99.5960 99.6074
AC

X_Chest 99.6223 99.6047

CT_Abdomen (  ) 99.6196 99.6067
CT_Hand 99.6238 99.6131
CT_Head 99.5826 99.6166
CT_Paranasal_sinus 99.6162 99.6246
MR_Brain 99.6044 99.5952
MR_Cervical_Vertibra 99.6093 99.5887
MR_Knee 99.5944 99.6074
X_Chest 99.5948 99.6074
 ACCEPTED MANUSCRIPT

(a) (b)

T
IP
Figure.10 (a) plain image sensitivity test for a set of random standard and medical images; (b) key
sensitivity test for 100 different dynamic key.

CR
5.4. Time consuming

Execution-time is also an important factor, with respect to security level. The duration of the proposed

US
cipher algorithm is evaluated and compared with those proposals presented in [Ref1,Ref27,Ref30-Ref32],
under grayscale images of different sizes. The implementation was realized using C programming language
and the experiment was performed using Intel(R) Core (TM) i7-CPU of 3GHz with 8GB of memory. The
AN
obtained results are given in Table 22, where the superiority of our proposal, in term of its effectiveness
and speed is demonstrated.
M

Table 22. The execution-time performance test.

Image size Encryption time (ms)

ED

Ref.[1] Ref.[27] Ref.[30] Ref[31] Ref[32] Proposed

256x256 178 109 7641 189 569 48
512x512 663 390 34768 758 2251 139
1024x1024 3142 1482 151709 3097 8986 481
PT

7. Conclusion
CE

To address both the unsuitability problem of existing bit-level proposals for image encryption, where
the specific features of medical images in not considered, and the efficiency problem encountered by
several image ciphers that have been suggested in the scientific literature, a new approach is introduced in
AC

this paper for efficient and secure image content preservation, this method it well-suited for both standard
and medical images, and it is based on the iterative application of two modules: chaotic confusion and pixel
diffusion, and controlled by means of an improved 1D chaotic system (i.e., Logistic Tent System (LTS)), in
which its initial conditions are dynamically produced and relied on both the external secret key and the
SHA-256 hash value of the plain image, that lead to generate one-time key-streams, promote the sensitivity
to minor alterations applied to plain image, and hence withstand to plain/chosen plain image attacks. In the
confusion module, a novel nonlinear bit-shuffling and circular shifting technique is handled, to further meet
the bit balancing effect, pixel value mixing effect, and certain diffusion mechanism. In the diffusion phase,
every confused pixel is diffused by means of an improved XOR operation (eXOR), aiming to elevate the
sensitivity to plain image, and accelerate the diffusion mechanism of the whole cipher algorithm. As the
diffusion mechanism with respect to pixel level mixing are introduced by the two modules, a good trade-
 ACCEPTED MANUSCRIPT

off between computational performance and sufficient security, can be achieved with just one encryption
round. The extensive analysis and tests have validated the robustness of our method against the commonly
known cryptographic attacks, and its effectiveness in term of security level and fast encryption time.

References

[1]. ZHOU, Yicong, BAO, Long, et CHEN, CL Philip. A new 1D chaotic system for image encryption. Signal
processing, 2014, vol. 97, p. 172-182.
[2]. LIU, Hongjun et WANG, Xingyuan. Triple-image encryption scheme based on one-time key stream generated
by chaos and plain images. Journal of Systems and Software, 2013, vol. 86, no 3, p. 826-834.
[3]. CHAPANERI, Santosh, CHAPANERI, Radhika, et SARODE, Tanuja. Evaluation of Chaotic Map Lattice systems for
image encryption. In : Circuits, Systems, Communication and Information Technology Applications (CSCITA),
2014 International Conference on. IEEE, 2014. p. 59-64.

T
[4]. WANG, Xingyuan et ZHANG, Hui-li. A color image encryption with heterogeneous bit-permutation and
correlated chaos. Optics Communications, 2015, vol. 342, p. 51-60.

IP
[5]. ZHU, Zhi-liang, ZHANG, Wei, WONG, Kwok-wo, et al. A chaos-based symmetric image encryption scheme
using a bit-level permutation. Information Sciences, 2011, vol. 181, no 6, p. 1171-1186.
[6]. CHEN, Jun-xin, ZHU, Zhi-liang, FU, Chong, et al. An image encryption scheme using nonlinear inter-pixel

CR
computing and swapping based permutation approach. Communications in Nonlinear Science and Numerical
Simulation, 2015, vol. 23, no 1, p. 294-310.
[7]. ZHANG, Wei, WONG, Kwok-wo, YU, Hai, et al. A symmetric color image encryption algorithm using the
intrinsic features of bit distributions. Communications in Nonlinear Science and Numerical Simulation, 2013,

[8].
vol. 18, no 3, p. 584-600.
US
CHEN, Jun-xin, ZHU, Zhi-liang, FU, Chong, et al. A fast chaos-based image encryption scheme with a dynamic
state variables selection mechanism. Communications in Nonlinear Science and Numerical Simulation, 2015,
vol. 20, no 3, p. 846-860.
AN
[9]. FRANÇOIS, Michael, GROSGES, Thomas, BARCHIESI, Dominique, et al. Image encryption algorithm based on a
chaotic iterative process. Applied Mathematics, 2012, vol. 3, no 12, p. 1910.
[10]. SHANNON, Claude Elwood. A mathematical theory of communication. ACM SIGMOBILE Mobile Computing
and Communications Review, 2001, vol. 5, no 1, p. 3-55.
M

[11]. EL ASSAD, Safwan et FARAJALLAH, Mousa. A new chaos-based image encryption system. Signal Processing:
Image Communication, 2016, vol. 41, p. 144-157.
[12]. DEL REY, A. Martín, SÁNCHEZ, G. Rodríguez, et DE LA VILLA CUENCA, A. A protocol to encrypt digital images
ED

using chaotic maps and memory cellular automata. Logic Journal of IGPL, 2015, vol. 23, no 3, p. 485-494.
[13]. BAKHSHANDEH, Atieh et ESLAMI, Ziba. An authenticated image encryption scheme based on chaotic maps
and memory cellular automata. Optics and Lasers in Engineering, 2013, vol. 51, no 6, p. 665-673.
[14]. FU, Chong, MENG, Wei-hong, ZHAN, Yong-feng, et al. An efficient and secure medical image protection
PT

scheme based on chaotic maps. Computers in biology and medicine, 2013, vol. 43, no 8, p. 1000-1010.
[15]. KHANZADI, Himan, ESHGHI, Mohammad, et BORUJENI, Shahram Etemadi. Image encryption using random bit
sequence based on chaotic maps. Arabian Journal for Science and engineering, 2014, vol. 39, no 2, p. 1039-
1047.
CE

[16]. NOROUZI, Benyamin, SEYEDZADEH, Seyed Mohammad, MIRZAKUCHAKI, Sattar, et al. A novel image
encryption based on row-column, masking and main diffusion processes with hyper chaos. Multimedia Tools
and Applications, 2015, vol. 74, no 3, p. 781-811.
YAVUZ, Erdem, YAZICI, Rifat, KASAPBAŞI, Mustafa Cem, et al. A chaos-based image encryption algorithm with
AC

[17].
simple logical functions. Computers & Electrical Engineering, 2015.
[18]. FRIDRICH, Jiri. Symmetric ciphers based on two-dimensional chaotic maps. International Journal of Bifurcation
and chaos, 1998, vol. 8, no 06, p. 1259-1284.
[19]. CHEN, Guanrong, MAO, Yaobin, et CHUI, Charles K. A symmetric image encryption scheme based on 3D
chaotic cat maps. Chaos, Solitons & Fractals, 2004, vol. 21, no 3, p. 749-761.
[20]. KANSO, A. et GHEBLEH, M. An efficient and robust image encryption scheme for medical applications.
Communications in Nonlinear Science and Numerical Simulation, 2015, vol. 24, no 1, p. 98-116.
[21]. WONG, Kwok-Wo, KWOK, Bernie Sin-Hung, et YUEN, Ching-Hung. An efficient diffusion approach for chaos-
based image encryption. Chaos, Solitons & Fractals, 2009, vol. 41, no 5, p. 2652-2663.
[22]. FU, Chong, CHEN, Jun-jie, ZOU, Hao, et al. A chaos-based digital image encryption scheme with an improved
diffusion strategy. Optics Express, 2012, vol. 20, no 3, p. 2363-2378.
[23]. ZHANG, Xuanping et ZHAO, Zhongmeng. Chaos-based image encryption with total shuffling and bidirectional
diffusion. Nonlinear Dynamics, 2014, vol. 75, no 1-2, p. 319-330.
 ACCEPTED MANUSCRIPT

[24]. CHEN, Jun-xin, ZHU, Zhi-liang, et YU, Hai. A fast chaos-based symmetric image cryptosystem with an improved
diffusion scheme. Optik-International Journal for Light and Electron Optics, 2014, vol. 125, no 11, p. 2472-
2478.
[25]. WANG, Yong, WONG, Kwok-Wo, LIAO, Xiaofeng, et al. A new chaos-based fast image encryption algorithm.
Applied soft computing, 2011, vol. 11, no 1, p. 514-522.
[26]. WANG, Xingyuan et GUO, Kang. A new image alternate encryption algorithm based on chaotic map.
Nonlinear dynamics, 2014, vol. 76, no 4, p. 1943-1950.
[27]. SOUYAH, Amina et FARAOUN, Kamel Mohamed. An image encryption scheme combining chaos-memory
cellular automata and weighted histogram. Nonlinear Dynamics, 2016, vol. 86, no 1, p. 639-653.
[28]. WANG, Xingyuan et XU, Dahai. A novel image encryption scheme based on Brownian motion and PWLCM
chaotic system. Nonlinear Dynamics, 2014, vol. 75, no 1-2, p. 345-353.
[29]. ZHANG, Ying-Qian et WANG, Xing-Yuan. A symmetric image encryption algorithm based on mixed linear–
nonlinear coupled map lattice. Information Sciences, 2014, vol. 273, p. 329-351.
[30]. WU, Yue, YANG, Gelan, JIN, Huixia, et al. Image encryption using the two-dimensional logistic chaotic map.

T
Journal of Electronic Imaging, 2012, vol. 21, no 1, p. 013014-1-013014-15.
[31]. MOHAMED, Faraoun Kamel. A parallel block-based encryption schema for digital images using reversible

IP
cellular automata. Engineering Science and Technology, an International Journal, 2014, vol. 17, no 2, p. 85-94.
[32]. LIAO, Xiaofeng, LAI, Shiyue, et ZHOU, Qing. A novel image encryption algorithm based on self-adaptive wave

CR
transmission. Signal Processing, 2010, vol. 90, no 9, p. 2714-2722.
[33]. ZHANG, Guoji et LIU, Qing. A novel image encryption method based on total shuffling scheme. Optics
Communications, 2011, vol. 284, no 12, p. 2775-2780.
[34]. BORUJENI, Shahram Etemadi et ESHGHI, Mohammad. Chaotic image encryption system using phase-

[35]. US
magnitude transformation and pixel substitution. Telecommunication Systems, 2013, vol. 52, no 2, p. 525-
537.
SONG, Chun-Yan, QIAO, Yu-Long, et ZHANG, Xing-Zhou. An image encryption scheme based on new
spatiotemporal chaos. Optik-International Journal for Light and Electron Optics, 2013, vol. 124, no 18, p.
AN
3329-3334.
[36]. LI, Yueping, WANG, Chunhua, et CHEN, Hua. A hyper-chaos-based image encryption algorithm using pixel-
level permutation and bit-level permutation. Optics and Lasers in Engineering, 2017, vol. 90, p. 238-246.
[37]. XU, Lu, GOU, Xu, LI, Zhi, et al. A novel chaotic image encryption algorithm using block scrambling and dynamic
index based diffusion. Optics and Lasers in Engineering, 2017, vol. 91, p. 41-52.
M

[38]. CHEN, Jun-xin, ZHU, Zhi-liang, FU, Chong, et al. An efficient image encryption scheme using gray code based
permutation approach. Optics and Lasers in Engineering, 2015, vol. 67, p. 191-204.
[39]. HUA, Zhongyun et ZHOU, Yicong. Image encryption using 2D Logistic-adjusted-Sine map. Information
ED

Sciences, 2016, vol. 339, p. 237-253.

PT
CE
AC