Hindawi

Security and Communication Networks

Volume 2022, Article ID 5680357, 16 pages
https://doi.org/10.1155/2022/5680357

Research Article
An Efficient Lightweight Image Encryption Scheme
Using Multichaos

Asad Ullah,1 Awais Aziz Shah,1,2 Jan Sher Khan,3 Mazhar Sajjad,4 Wadii Boulila ,5,6
Akif Akgul ,7 Junaid Masood,8 Fuad A. Ghaleb ,9,10 Syed Aziz Shah,10
and Jawad Ahmad 11
1
School of Computing, Engineering and Physical Sciences, University of the West of Scotland, Paisely PA12BE, UK
2
Department of Electrical and Informational Engineering (DEI), Polytechnic University of Bari, Bari, Italy
3
Department of Electrical and Electronics Engineering, University of Gaziantep, Gaziantep, Turkey
4
Department of Computer Science, Comsats University, Islamabad 45550, Pakistan
5
RIADI Laboratory, University of Manouba, Manouba, Tunisia
6
Robotics and Internet of Things Laboratory, Prince Sultan University, Riyadh 12435, Saudi Arabia
7
Department of Computer Engineering, Faculty of Engineering, Hitit University, Corum 19030, Turkey
8
Department of Computer Science, IQRA National University, Peshawar, Pakistan
9
Department of Computer and Electronic Engineering, Sana’a Community College, Sana’a, Yemen
10
School of Computing, Faculty of Engineering, Universiti Teknologi Malaysia, Johor Bahru 81310, Malaysia
11
Research Centre for Intelligent Healthcare, Coventry University, Coventry, UK

Correspondence should be addressed to Fuad A. Ghaleb; [email protected]

Received 28 July 2021; Revised 18 August 2021; Accepted 27 July 2022; Published 14 October 2022

Academic Editor: Farhan Ullah

Copyright © 2022 Asad Ullah 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.

With an immense increase in Internet multimedia applications over the past few years, digital content such as digital images are
stored and shared over global networks, the probability for information leakage and illegal modifications to the digital content is at
high risk. These digital images are transferred using the network bandwidth; therefore, secure encryption schemes facilitate both
information security and bandwidth issues. Hence, a state-of-the-art lightweight information security methodology is required to
address this challenge. The main objective of this work is to develop a lightweight nonlinear mechanism for digital image security
using chaos theory. The proposed scheme starts by changing a plain image into an encrypted image to improve its security. A block
cipher, using lightweight chaos, has been added to achieve this objective for digital image security. We utilized multiple chaotic
maps to generate random keys for each channel. Also, Arnold cat map and chaotic gingerbread map are used to add confusion and
diffusion. During the permutation stage, image pixels are permuted, while in diffusion stage, pixels are distorted utilizing
gingerbread map to add more security. The proposed scheme has been validated using different security parameter tests such as
correlation coefficient tests (CC), whose results have been observed closer to zero and information entropy (IE) value is 7.99,
respectively, which is almost equal to the ideal value of 8. Moreover, number of pixels changing rate (NPCR) obtained value is
higher than 99.50%, while the unified average changing intensity (UACI) is 33.33. Other parameters such as mean absolute error
(MAE), mean square error (MSE), lower value of peak to signal noise ratio (PSNR), structural content (SC), maximum difference
(MD), average difference (AD), normalized cross-correlation (NCC), and histogram analysis (HA) is tested. The computed values
of the proposed scheme are better. The achieved results after comparison with existing schemes highlight that the proposed
scheme is highly secure, lightweight, and feasible for real-time communications.
2 Security and Communication Networks

1. Introduction initial value sensitivity. One advantage of employing

chaos-based image encryption is that it offers higher
There has been a rapid growth in the computers, Internet, security even though the mathematical complexity in-
social media, and communications during the last decade. volved is much lower [18]. However, such systems’
The information revolution consists of digital media narrow and discontinuous range causes massive differ-
transmitted from billions of users and devices globally ences in chaotic properties due to minor disturbances
using network bandwidth. This increase in digital media such as noise factors. Two major factors that controls the
transmission poses numerous challenges in terms of performance of chaos-based encryption algorithms for
privacy and security as hacking incidents have occurred imaging. The first such factor is based upon the ro-
in the last decade that consists of hacking the content for bustness of the encryption algorithm against cryptanal-
information leakage and modifications [1–4]. More re- ysis. The second factor is based upon the real-time
cently, numerous cryptographic methods have been performance of chaotic maps. If the obtained perfor-
proposed to enable secure image encryption. However, mance is lower, this means that the achieved security
these methods often have an ordinary structure that robustness and efficiency of chaotic maps is lower. When
performs the permutation and diffusion stages of de- both these factors are combined to attempt better results,
cryption. Most of these algorithms deal with issues such many new challenges and difficulties also arise. So, their
as lack of robustness and security. Additionally, these balance is essential in this regard.
encryption techniques were unsuitable for images with This paper presents an efficient and lightweight image
certain features, i.e., huge capacity, high redundancy, and encryption technique based on multiple chaos algo-
high correlation among pixels. In addition to these rithms. The work includes designing a novel and efficient
characteristics, images also tend to be much larger than information safety scheme based on various chaotic
other media that might require decryption. These en- maps. Numerous encryption techniques are proposed
cryption plans require additional operations on com- and analyzed for their computational speed, reliability,
pressed image data, demanding long computational time robustness, and nonlinearity, which has been found ex-
and high computing power. This translates to low en- ceptional. Our primary research objective is to achieve a
cryption and decryption speeds in real-time communi- high level of security. The security of a given system
cations, and images may present significant latency [5, 6]. depends upon ensuring its cryptographic services. Several
With recent advancements in social media platforms, services of protection are provided by the International
many images are transmitted over the web daily. Securing Telecommunication and Union’s Telecommunication
such images against unauthorized access becomes consid- Standardization. We investigate the chaotic properties,
erably more critical for authorized people and various fields including topology mixing, strange attractor, ergodicity,
to confront their unique security issues. Biology, military randomness, and dependence on its initial condition.
sciences, banking, and online shopping are just a few fields to After that, we evaluated the proposed cryptosystem using
name. In these cases, information transmitted as or via different tests, such as histogram analysis (HA), corre-
images is very different from information transmitted as or lation coefficient (CC), the mean absolute error (MAE),
via text [7]. Images deal with greater data redundancy and differential attacks analysis (DAA), number of pixels
more excellent information coverage while having a higher changing rate (NPCR), unified average changing intensity
correlation between adjacent pixels. They also require real- (UACI), the mean square error (MSE), peak to signal
time and robust security for communication purpose. Thus, noise ratio (PSNR), information entropy (IE), structural
numerous techniques are used to secure confidential image content (SC), normalized cross-correlation (NCC),
data transmitted over an insecure channel [8–12]. Data maximum difference (MD), and average difference (AD).
sharing is increased using social networks that are expanding The comparison of proposed algorithms results in
in size and traffic, which has altered the underlying infra- existing algorithms for all tests. The results highlight that
structure of communication. calculated CC value is closer to zero, and IE value is 7.99,
Today, much effort and research have created a com- respectively, which is almost equal to the ideal value of 8.
plicated and robust cryptosystem providing the best possible Moreover, NPCR is higher than 99.50%, while the UACI
security. Such a system uses several highly desired crypto- is 33.33. Other parameters such as MAE, MSE, lower
graphic properties, such as randomness, ergodicity, and value of PSNR, SC, MD, AD, and NCC showed better
sensitivity to original values [8]. result in the proposed scheme. The offered scheme is
As technology’s scope ascends, new methods are compared to the existing schemes. It is concluded that the
developed, including special techniques such as deoxy- proposed scheme is highly secure, lightweight, and fea-
ribonucleic acid (DNA) coding [13], quantum theory, sible for real-time communications.
compression, and chaos theory. Though all of these have The rest of the article has been organized as follows:
their advantages, chaos theory, in particular, has played a Section 2 gives an overview and relation between chaos and
critical role in encryption research over the last 20 years cryptography, Section 3 deeply describes the proposed
[14–17]. By definition, the chaos theory uses mathematics methodology of the lightweight image encryption scheme.
and physics to deal with the behavior of NLDSs, i.e., The statistical analysis and experimental results are dis-
nonlinear dynamical systems. The typical characteristics cussed in Section 4 and finally, Section 5 concludes the
of chaotic systems are ergodicity, unpredictability, and paper.
Security and Communication Networks 3

2. Chaos and Cryptography of chaos. A long-term progression that is continuous, erratic,

and fulfills mathematical benchmarks will display chaos
It has been made clear that data transmitted over any public phenomena. The features of a chaotic system can be de-
network is susceptible to malicious attacks and might be- scribed as a set of properties that measure mathematical
come a target to be a break. For the sole purpose of pro- principles that are used to describe chaos. The most notable
tecting communications in general, several encryption of these characteristics include the following:
techniques have been suggested. This has emphasized dy-
namic cryptosystems, such as chaos systems, that transmit (1) Nonlinearity: The output is changed unpropor-
plain data into unintelligible cipher data. Sensitivity is one of tionally concerning input.
the unique features of such systems, i.e., any minor change in (2) Deterministic: Every state of the system must follow
the original condition will result in drastically random some deterministic fundamental rules.
output formulas such as the Lyapunov exponent help cal- (3) Sensitivity to initial conditions: Slight deviation in its
culate the parameters of a map. Randomness, initial sen- early state can lead to a dissimilar performance in its
sitivity, and periodicity are more beneficial properties that last state.
cryptographer use to design well-built cryptographic algo- (4) Continued irregularity in the system’s actions: The
rithms that provide security against unauthorized users. framework of chaotic systems is based upon a secret
Among these systems, chaotic systems stand out due to their order combined with an infinite number of unstable
random output and other unique features [19]. As chaotic random designs. This unseen direction forms the
systems can perform without any dependence on parameters structure of irregular, chaotic systems.
[19], they demonstrate in-determinism, thus helping other
creators to design better algorithms. A chaotic output can be (5) Long-term prediction: It can only be comprehended
determined using its initial values for the system, making the with limited accuracy as chaos is sensitive to the
chaotic more deterministic. When combined, these features initial state.
help to diversify cryptosystems meticulously. When
equipped with diffusion and confusion, a nonlinear system 2.2. Elementary Possessions of Chaotic Systems. Chaos has
can effectively encounter cryptanalysis. Due to this robust been measured in various laboratories, and natural systems
infrastructure involved in chaotic cryptosystems, these [20, 21], covering many engineering and scientific areas,
processes are now being employed in other fields such as including fields such as biology, physics, ecology, meteorol-
biological sciences, chemistry, and physics to build a new set ogy, economics, and computer science, among others. All the
of hybrid systems to achieve better security. Figure 1 il- techniques above provide valuable assets, which create certain
lustrates the basic schematic chart of chaos-based encryption unpredictability and a complex system. Systems that display
schemes. such erratic and dynamic behavior over time are explained by
In an NLDS, i.e., a nonlinear dynamical system, initial chaos theory. For a broader overview, the concerned reader is
conditions significantly impact chaos, which seems to work directed to (Robinson, 1995). Generally, these systems are
upon and show pseudo-random behavior. Consider the deterministic, and they follow a specific set of laws of evo-
stability of the system as an essential parameter when lution. Unfortunately, it must be said that chaos only occurs in
considering Lyapunov exponents. When the observer is a certain deterministic NLDS (nonregular linear dynamic
aware of the system’s initial conditions, the observer seems systems). The most relevant of these criteria are dynamic
to understand the system output. However, if the prelimi- instability, topological mixing, aperiodicity, and ergodicity:
nary requirements are unknown, the overall production Now let us examine the interaction and connection of
seems to be highly erratic and random. When the source chaos with cryptography properties of diffusion and con-
owner knows the pseudo-random order of the data, then the fusion (Shannon, 1949). Suggesting to review chaotic sys-
plain text can be easily substituted and diffused to be pro- tems features, like ergodicity, topology mixing property, and
tected against malicious attacks and invasion. Several data auto-similarity that are directly connected to the confusion
formats can also be utilized in telecommunication channels phase of chaotic systems. The dynamics, or dynamical be-
that require protection. havior in the chaotic output attractor, is concerned with
nonperiodic orbits that produce identical patterns. The
output patterns can be utilized to mask clear messages using
2.1. Fundamental Properties of Chaotic System. Many tech- substitution-like techniques. On the other hand, diffusion is
nical and industrial areas within natural structures witness directly connected to the system’s sensitive parameters and
the chaos. These areas often exhibit well-defined possessions initial conditions. Small changes in control parameters give
marking them as complicated and highly volatile. The chaos rise to avalanche effects and produce random outputs.
theory is concerned with situations that move through time As a conclusion, it would be best to list the benefits of
toward a specific type of dynamic action. Multiple authors chaotic cryptosystems that are being used and to offer strong
across various fields have discussed the chaos theory’s security in cryptography. Firstly, chaotic systems seem to
mathematical background due to its wide variety of appli- happen on their own and can be used for lightweight security
cations. In broad terms, some specific procedures are fol- in cryptography. These maps may show nonlinear behav-
lowed by these schemes for improvement. Individual iours that are used to make communication more secure.
nonlinear deterministic systems are usually potential sources These systems have the advantage of having simple,
4 Security and Communication Networks

Receiver
Sender end
end
Cryptographic algorithm

Chaos based
Shared secret
cryptographic
key
algorithm

Plain image Encrypted

image

Figure 1: Basic schematic chart of chaos-based encryption process.

Sender Plain image Receiver Encrypted

image

Encryption process Decryption process

Insecure channel
Encryption Private/Public Private/Public Decryption
algorithm key key algorithm

Figure 2: The process of encryption and decryption.

deterministic algorithms that are easy to use. Three basic In this section, the design of such a cryptosystem is
ideas can be used to build a security infrastructure: confi- discussed. This discussion will include all requirements that
dentiality, integrity, and authentication. Confidentiality are mandatory for creating a truly secure cryptosystem. The
means that only people who are allowed to can see the data proposed system is then investigated using several tests to
and people who are not should not. Integrity makes sure that secure the image and its transmission remains. During this
the information available has not been changed. Availability process, we evaluate the performance of the proposed al-
means that there are resources that can be used. gorithm. We have conducted statistical tests including HA,
CC tests between pixels along three primary axes, i.e.,
horizontal, diagonal, and vertical for each channel. MSE,
3. The Proposed Methodology of Lightweight PSNR, IE against each channel, sensitivity analysis which
Image Encryption Scheme interpolates NPCR, and UACI. Other tests such as an AD,
MD, SC, and NCC are performed. Our experiments enabled
Figure 2 shows the general process of encryption and de- us to instantly examine output bit pixel, input bit pixel, and
cryption. Lightweight cryptosystems that operate based on their behavior after investigation. Our primary goal is to
chaos principles are designed by utilizing confusion and create algorithms that include confusion and diffusion
diffusion characteristics noted by Claude Shannon [22], as (randomness) since these are two essential properties for
discussed earlier in Section 2. Such lightweight encryption optimal security service.
systems work by using several chaotic maps to build hybrid
systems that bring together several initial systems’ best
characteristics. The suggested method is then subjected to 3.1. Utilized Chaotic Maps. To help develop and implement
robust testing parameters compared with other existing encryption algorithms, general chaotic maps will be defined
techniques to determine its efficiency and efficacy. and discussed.
Security and Communication Networks 5

3.1.1. Arnold Scrambling Cat Map. Arnold’s method is a (4) Next, two random and arbitrary chaotic matrices are
standard technique for scrambling an image. This method generated using a gingerbread man chaotic map.
is secure because of its high computational rate and quick These two random matrices are bitwise XORed with
processing time. Moreover, this method facilitates data each other to develop diffused channels, and the
scrambling in an array as a data stream using a chaotic resultant matrix is bitwise XORed with the red
Arnold transformation. Arnold’s transformation chaotic permuted channel.
map can be defined as follows: (5) To make the scheme ciphertext dependent, the re-
x′ 1 1 x sultant red channel diffused channel is bitwise
⎣
⎡ ⎤⎦ � 􏼠 􏼡􏼢 􏼣mod1, (1) XORed with a green permuted channel. Their result
y′ 1 2 y matrix is bitwise XORed with a blue permuted
where x and y ∈ 0, 1. The aforementioned formula of the channel.
Arnold chaotic map is for unit square and can be extended to (6) The multi-layered cryptosystem then produced three
multiple rows and columns depends upon the pixels of the sets of encrypted images by applying different cha-
image to be encrypted having a size of N1(i,j) × N2(i,j) In this otic maps.
case, then the above matrix can be extended to the finite (7) For the final step, the cat command module is
number of pixels. employed to combine the three encrypted channels
x′ 1 1 x and generate the final ciphertext image.
⎣
⎡ ⎤⎦ � 􏼠 􏼡􏼢 􏼣modN, (2)
y′ 1 2 y
where x and y ∈ {0, 1, 2, 3, . . . , N − 1} are the original im- 3.2.2. Steps Involved for Image Decryption
ages pixels scrambled and transformed to new positions, i.e., (1) Initially, the encrypted image is divided into its
􏼈x′ , y′ 􏼉. respective three parts (R, G, B).
(2) The green and blue cipher channels are bitwise
3.1.2. Gingerbread Man Chaotic Map. The gingerbread man XORed to get a blue permuted channel. Then the red
is a 2D chaotic map that is one of the most widely utilized and blue cipher channels are bitwise XORed to
choices in chaotic, random sequencing. However, it is also obtain a blue permuted channel.
essential to have a robust cryptosystem to satisfy all the (3) Two random matrices are generated using a gin-
designed algorithm’s security needs when using this option. gerbread man chaotic map. These two random
The formula for this 2D piecewise chaotic linear map is as matrices are bitwise XORed with each other, and the
follows: resultant matrix is bitwise XORed with the red cipher
􏼌􏼌 􏼌􏼌
xn+1 � 1 − yn + 􏼌􏼌xn 􏼌,􏼌 channel to produce a red permuted layer.
(3) (4) The 2D Arnold chaotic map is then used to reiterate
yn+1 � xn .
the three different layers row- and column-wise from
maximum permuted (iterated) state to the original
condition.
3.2. Image Encryption Process
(5) Three plain channels are recovered after employing
3.2.1. Steps Involved in the Process of Image Encryption. Arnold cat scrambling for the reiteration process.
The step wise flowchart of the proposed scheme is illustrated (6) The plain grey channels are combined using cat
in Figure 3. The pseudo code of the image encryption process command
is illustrated in Algorithm 1. Let us consider a baboon test (7) The decrypted full-colored image is retrieved.
image measuring 256 × 256 × 3 pixels that will be encrypted
in the following process:
(1) First, this test image is fragmented into three 4. Statistical Analysis and Experimental Results
channels (red, green, and blue) with an image size of
Experiments are conducted utilizing multiple plain im-
256 × 256 pixels.
ages to demonstrate the effectiveness of the proposed
(2) To make the scheme plaintext dependent, each layer scheme. A good encryption strategy should be robust to
is passed through SHA-512. Then, the utilized cha- address the needs engendered by cryptanalysis, statistics,
otic maps such as logistic map, Gaussian map, and and brute force attacks, respectively. Along with these
Chebyshev chaotic map initial conditions (keys) are considerations, the goal is to analyze security investi-
computed using the calculated hash values. gations to demonstrate the effectiveness of the proposed
(3) In this phase, the original image is permuted approach against widely-recognized attacks. To illustrate
(confused) using 2D Arnold scrambling. The the strength of our developed methodology, a statistical
plaintext channels are permuted row- and column- investigation has been conducted by creating a schematic
wise from their initial position to the maximum histogram and correlating the nearby pixels in both plain
iterative state. and cipher images.
6 Security and Communication Networks

SHA-512 SHA-512 SHA-512

Key 1 Key 2 Key 3

Confusion: Arnold cat scrambling (Permutation)

Row-wise Row-wise Row-wise

Column-wise Column-wise Column-wise

Diffusion: Gingerbread man chaotic map (Substitution)

Matrix 1

Red diffused Green diffused Blue diffused

channel channel channel
Matrix 2

Diffused
Matrix

Final Ciphertext Image

Figure 3: Flow chart of the proposed scheme.

4.1. Histogram Analysis (HA). It is widely used as a tool for uniform, while in the case of a plain image, these pixels
analyzing the security of encrypted images. The original are bumpy. The bumpy pixels demonstrate that anyone
content in the plain image must be secure. This test can access the pixel information easily. High peaks of
explains that the grey-level pixel value should be equally pixels reveal that maximum information lies at those
distributed in its defined range depending on the image. points while low peaks indicate that minimum infor-
In the cipher image, the distribution of pixels is always mation lies at this point. A secure image histogram
Security and Communication Networks 7

Inputs: Plaintext image P, its row and column numbers row, col,
Outputs: Ciphertext image C
(1) for i � 1 to row
(2) for j � 1 to col
(3) R, G, B←P
(4) Key1, Key2, Key3←SHA256(R, G, B)
(5) R, G, B←Row- and column −wisepermutation
(6) Permuted(R, G, B)←bit − wiseXORed
(7) Endfor
(8) Endfor
(9) C←Diffused(R, G, B)

ALGORITHM 1: Image encryption.

(a) (b) (c)

8000 3000
2500
6000
2000
4000 1500
1000
2000
500
0 0
0 50 100 150 200 250 0 50 100 150 200 250

(d) (e) (f)

4000
3000
3000
2000
2000
1000 1000

0 0
0 50 100 150 200 250 0 50 100 150 200 250

(g) (h) (i)

Figure 4: Continued.
8 Security and Communication Networks

(j) (k) (l)

2500
7000
6000 2000
5000
1500
4000
3000 1000
2000
500
1000
0 0
0 50 100 150 200 250 0 50 100 150 200 250

(m) (n)
2500 2500

2000 2000

1500 1500

1000 1000

500 500

0 0
0 50 100 150 200 250 0 50 100 150 200 250

(o) (p)

Figure 4: Encryption, decryption, and histogram results of the proposed scheme. (a, b, c, and d) Plaintext Pepper, plaintext red layer Pepper,
plaintext green layer Pepper, and plaintext blue layer Pepper images, respectively; (e, f, g, and h) corresponding histograms; (i, j, k, and l)
corresponding ciphertext images; (m, n, o, and p) ciphertext histograms.

creates a uniform pattern with minimal peaks and valleys, encrypted) is low or close to zero, the proposed system is
making it difficult for hackers to read the actual content secure. In comparison, in the case of highly similar pixels
due to pixel uniformity. The histogram shows the sta- between the two images, this demonstrates that the
tistical quality of various images. It also discloses how system is insecure. The defined range of correlation falls
arbitrary numbers created from chaotic maps, such as within [−1 1]. The high correlation of pixels between the
white noise, are uniformly dispersed. The histograms of two images reveals that the system is susceptible to many
plain images and encrypted images are given in Figure 4. types of attacks and that malicious third parties can easily
detect an image’s original confidential content. In the
case of encrypted images, the pixels are scattered in its
4.2. Correlation Coefficient Analysis (CC). It is used to defined range, depicting that the proposed system is
determine the relationship between two variables. To secure. Figure 5 demonstrates the correlation plots for
discover the relationship between two variables, in our Pepper image. The test is applied on color and a channel-
case, CC analysis is used. CC finds the quality of wise image with an image size of 512 × 512 × 3 and 512 ×
encrypted image by analyzing pixels distribution [23]. 512 for three grey channels, respectively. Thus, one can
How effective a cryptosystem is can be measured by the confirm that in case of plaintext, the correlation among
encryption algorithm’s ability to mask all plain image the adjacent pixels is very much strong and the plots are
characteristics. Furthermore, it relies on an image’s diagonally condensed. While in case of ciphertext, the
randomness and how uncorrelated the cipher image is plots are highly scattered which confirms high dissimi-
[23–25]. If the CC for the two images (i.e., plain and larity and randomness generated using the presented
Security and Communication Networks 9

250 250 250

200 200 200

150 150 150

100 100 100

50 50 50

0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(a) (b) (c)
300 300 300
250 250 250
200 200 200
150 150 150
100 100 100
50 50 50
0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(d) (e) (f )
300 300 300
250 250 250
200 200 200
150 150 150
100 100 100
50 50 50
0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(g) (h) (i)
300 300 300

250 250 250

200 200 200

150 150 150

100 100 100

50 50 50

0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(j) (k) (l)
300 300 300

250 250 250

200 200 200

150 150 150

100 100 100

50 50 50

0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(m) (n) (o)
Figure 5: Continued.
10 Security and Communication Networks

300 300 300

250 250 250

200 200 200

150 150 150

100 100 100

50 50 50

0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(p) (q) (r)
250 250 250

200 200 200

150 150 150

100 100 100

50 50 50

0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(s) (t) (u)
300 300 300

250 250 250

200 200 200

150 150 150

100 100 100

50 50 50

0 0 0
0 50 100 150 200 250 0 50 100 150 200 250 0 50 100 150 200 250
(v) (w) (x)

Figure 5: Pepper image correlation plots: (a, b, and c) plaintext Pepper correlation plots in horizontal, diagonal and vertical direction while
row (1) (d, e, and f ) corresponding ciphertext plots; (g, h, and i) plaintext Pepper red layer correlation plots in horizontal, diagonal and
vertical direction; (j, k, and l) corresponding ciphertext plots; (m, n, and o) plaintext Pepper green layer correlation plots in horizontal,
diagonal and vertical direction; (p, q, and r) corresponding ciphertext plots; (s, t, and u) plaintext Pepper blue layer correlation plots in
horizontal, diagonal and vertical direction; (v, w, and x) corresponding ciphertext plots.

Table 1: Correlation coefficient values for Pepper image of size 512 × 512.
Plaintext image Encrypted image
Directions Directions
HC DC VC HC DC VC
Ideal 1 1 1 −1 −1 −1
Pepper 0.9635 0.9564 0.9663 −0.0033 0.0011 0.007
Ref. [28] — — — 0.0075 0.0012 0.0049
Ref. [29] — — — 0.0005 0.0008 0.0011
Ref. [30] — — — 0.0117 0.0026 0.0010
Ref. [31] — — — 0.0043 0.0054 0.0072
Ref. [32] — — — 0.0108 0.0181 0.0061
Ref. [33] — — — 0.0032 0.0042 0.0018
Ref. [34] — — — 0.0204 −0.0174 0.0231
Ref. [35] — — — 0.0053 −0.0027 0.0016
HC : Horizontally correlated; DC : Diagonally correlated; VC : Vertically correlated.
Security and Communication Networks 11

scheme. These scattered plots verify the scheme resistant Table 2: MAE test values.
to an attack. Correlation can be mathematically described Image Size MAE
as [26, 27]
Pepper 512 × 512 80
Sxy Tiffany 512 × 512 182
r� . (4) Airplane 512 × 512 92
Sx Sy Ref. [36] pepper 256 × 256 74.93
Ref. [36] tiffany 256 × 256 94.36
where the system’s covariance is Sxy , while standard devi- Ref. [36] airplane 256 × 256 82.88
ations of random variables of Sx , Sy , are x, and y, respec-
tively. We applied the CC on the grey images of the same size
Table 3: AND UACI test values for different image (512 × 512)
as having a length of subsequently combined all the
channels.
encrypted grey channels for the colored image. We also
applied the CC test for the second time on colored images Image Test type R–C G–C B–C
having a size of respectively. The computed values of the CC Pepper 99.613% 99.613% 99.613%
UACI
are shown in Table 1. Pepper 33.01 33.39 33.75
Tiffany 99.626% 99.626% 99.626%
UACI
Tiffany 36.13 36.13 36.13
4.3. The Mean Absolute Error (MAE). MAE is one of the most
important criteria required to determine an image’s quality. Table 4: Computed and UACI values comparison with other
It is also used to test how resistant a method is toward schemes.
differential attacks. To analyze, let’s consider the image being
the total size of a test image. Let the C denotes the ciphered Image Test type Combined value
image, and P denotes the grey pixels of a plain image at the i Pepper 99.613%
UACI
th row and J th column. The equation used to find the mean Pepper 33.38
absolute error is Tiffany 99.626%
Tiffany 36.13
M−1 N−1􏼌 􏼌􏼌 Ref. [37] pepper 99.589%
1 􏼌
MAE � 􏽘 􏽘 􏼌􏼌􏼌C(i,j) − P(i,j) 􏼌􏼌􏼌. (5) Ref. [37] pepper
UACI
33.373
M × N i�0 j�0
Ref. [38] pepper 99.01%
UACI
Ref. [38] pepper 33.51
It is essential to obtain a higher value of MAE to validate Ref. [39] pepper 99.22%
the cryptosystem’s robustness. If the evaluated MAE value is UACI
Ref. [39] pepper 33.12
not higher than the existing scheme’s average values, then Ref. [34] pepper 99.61%
UACI
the system has not resisted the differential attack. In other Ref. [34] pepper 33.48
words, the higher the MAE, the stronger the security level of Ref. [40] pepper
UACI
99.60%
the designed cryptosystem. However, the resistant differ- Ref. [40] pepper 33.41
ential attack may still be fulfilled if the MAE computed value
is not higher than the existing cryptosystems’ values. The The techniques of UACI and NPCR are usually employed to
minimum value should be 75, which means that value higher measure the effectiveness of pixel alteration.
than 75 demonstrates a high-security level. Our findings
show that the proposed algorithm’s calculated values are
higher than 75: specifically, the estimated MAE value for the 4.4.1. Number of Pixels Changing Rate (NPCR). Several
pepper image is 80, while the MAE for the Airplane image is pixels’ change rate examines how many pixels transmute
92, both of which are significantly higher than 75. Mean- when a single pixel alters in the plaintext image. This
while, the MAE value of the “Tiffany” image over is two value reaches 99.60% in NPCR, showing us that the
times higher than either of those, as shown in Table 2. system has approached cautiously as it can easily defend
itself against plain text attack. Ideally, the value for NPCR
should be 100%. Tables 3 and 4 indicate that our proposed
4.4. Differential Attacks Analysis. The randomness of the system achieves a calculated value of 99.613% for the
proposed encryption scheme can be evaluated using per- encrypted pepper image across all three channels. In
mutation method. Permutation’s ability is to highlight even contrast, the Tiffany image’s calculated; value is 99.626%
small changes in plain images. To check the proposed sys- for its all respective channels, as shown in Table 3. The
tem’s sensitivity level, we change a single pixel in the plain proposed values and preexisting algorithms values are
image and then calculate its respective outcome. As dis- subsequently shown in Table 4, which compares the
cussed in previous sections, small changes in the plain image values we achieve to the values achievable by other
can lead to drastic changes in the encrypted image, making existing systems. Table 4 reveals that both images achieve
the connection between original and cipher images very a higher proposed calculated value than those achieved by
difficult and makes it more difficult for unauthorized third existing methods.
parties to decode them. Thus, the alteration of pixels in the Two cipher images we consider for this evaluation are C1
plain image allows us to expand the encryption algorithm. and C2. However, their source image differs from a single
12 Security and Communication Networks

pixel. The equation used to calculate the value is as follows Table 5: MSE and PSNR values for different channels of Pepper
[41]: and Tiffany image.

􏽐i,j Di,j Image Image channel Size MSE PSNR

� × 100, (6)
W×H Red 512 × 512 11117.33 7.70
Pepper Green 512 × 512 11222.95 7.66
where W × H denotes the total dimension of an image and Blue 512 × 512 7988.03 9.14
can be illustrated as follows: Red 256 × 256 177701.01 5.67
Case 1: Tiffany Green 256 × 256 13143.97 6.89
Blue 256 × 256 7347.84 9.50
D(i,j) � 0,
(7)
C1(i,j) � C2(i,j) . algorithm has excellent potential in the field of secure
communications.
Case 2:
D(i,j) � 1, 4.6. Peak to Signal Noise Ratio (PSNR). The image quality can
(8) be measured by utilizing the PSNR test. PSNR can be cal-
C1(i,j) ≠ C2(i,j) . culated as follows [34, 37–40]:
Imax2
PSNR � 10 log2 􏼒 􏼓. (11)
MSE
4.4.2. Unified Average Changing Intensity (UCAI). UACI is
used to test and compare the intermediate intensities of both Here I_max denotes the maximum number of pixels of the
plain and encrypted images. UACI can be defined as [42] test image. of pixels of the test image. PSNR is usually
follows: calculated in terms of decibels (dB). While MSE achieved
value is better to ensure robustness of the cryptosystem’s
1 C1(i,j) − C2(i,j)
UACI � 􏽘􏼢 􏼣100%. (9) security, PSNR should have a lower value for better data
W × H i,j 255 protection. The aforementioned Table 5 shows different
PSNR numbers for a set of additional test images. The av-
Here, C1(i,j) and C2(i,j) are the two ciphered images, while erage comes out to be 7, which satisfies the requirements of a
W × H denotes the total dimension of an image. We have robust cryptosystem.
considered a colored image that is divided into three layers
of the same size. The evaluated results for all the images are
higher than the average value of 33%, which indicates that 4.7. Information Entropy (IE). This test provides key points
the system has added a great deal of randomness into the regarding self-information. A report of any encryption
proposed cryptosystem. For the pepper image, the algo- channel requires entropy, and uncertainty itself relies upon
rithmic proposed value for those three channels (R, G, B) is entropy that is commonly known as randomness. IE can
33.01, 33.39, and 33.75, respectively, as tabulated in Table 3. explain the degree to which uncertainties are present in
The Tiffany image’s evaluated average values, and its three communication channels [43]. According to Claude Shan-
channels (R, G, B) are 36 for all three channels. Table 4 non, as early as 1949, information theory is based on the
compares the values shown with preexisting values from mathematics of data transfer and storage [28]. Today, in-
[34, 37–40], which are reported as 33.373, 33.51, 33.12, 33.48, formation theory is linked to communication systems,
and 33.41, respectively. These results indicate that the pre- cryptology, data compression, and many other similar topics
sented scheme has produced a higher security level against [44, 45]. Thus, entropy is the most significant characteristic
any type of external attack. of uncertainty and unpredictability since it demonstrates
irregularity in the behavior of both plain and encrypted data.
For the best-encrypted result, the cryptosystem must achieve
4.5. The Mean Square Error (MSE). MSE can be evaluated a value of entropy as near eight as possible. IE of any message
using the difference between the two images. When the pixels can be calculated as follows [46]:
of the plain and encrypted images are squared and then N−1
averaged, we get MSE that is calculated as follows [34, 37–40]: H(m) � 􏽘 p􏼐xj 􏼑logb p􏼐xj 􏼑. (12)
M N j�0
1 2
MSE � 􏽘 􏽘 􏼐P(i,j) − C(i,j) 􏼑 , (10)
M × N i�1 j�1 This explains the probability of any symbol to occur.
Suppose there is a truly random source generating about 28
where M × N represents the cumulative size of an image, symbols with an equal probability m will be
while i and j represent image rows and columns, p and C are � m1 . . . . . . . . . m28 , where every symbol is shown by 8 bits.
the plain and ciphered images at the ith row and jth column, Upon applying this for Eq. (10), an IE value of H(m) � 8 bits
respectively. A higher value of MSE shows that our suggested will be attained corresponding to a uniform random source.
scheme has a greater level of robustness. Table 5 contains As a whole, the ideal value is always greater than the IE of the
several values for MSE and PSNR, including an average layer actual source. This is because actual data rarely sends out
value of 10,000. Thus, we conclude that the proposed randomized information. In this case, as mentioned above, a
Security and Communication Networks 13

Table 6: Entropy values for different channels of Pepper and scheme includes substitution, permutation-like noise, and
Tiffany image. commotion in the plain image. SC cannot be approximated
Image Image channel Size Computed entropy as one if all of the shading images are advanced. SC cal-
Red 512 × 512 7.999
culated values for the proposed image encryption are il-
Pepper Green 512 × 512 7.999 lustrated in Table 8.
Blue 512 × 512 7.999
Red 256 × 256 7.999 4.9. Normalized Cross-Correlation (NCC). NCC is a sub-
dTiffany Green 256 × 256 7.999 stantial test used to find the sets of similarity in image sets,
Blue 256 × 256 7.999
particularly a plain image and its respective encryption of the
Red 512 × 512 7.999 same size. This test is widely utilized in image processing to
Splash Green 512 × 512 7.999 find the quality of an image as well. The range of NCC falls
Blue 512 × 512 7.999
between −1 and 1. Here −1 shows a strong correlation
Red 256 × 256 7.997 between the plain and encrypted images. Also, 1 shows that
Tiffany Green 256 × 256 7.997
the correlation between two images is not strong, also known
Blue 256 × 256 7.997
as a perfectly inverse correlation. Mathematically, NCC is
Red 512 × 512 7.999
defined as
Airplane Green 512 × 512 7.999
Blue 512 × 512 7.999 1 sum􏼐􏼂xn − meanx 􏼃 × 􏽨yn − meany 􏽩􏼑
NCC � × 􏽰���������� , (14)
N varx × vary
Table 7: Comparison of computed entropy values and pre- where M × N is the total size of an image and var shows
existing algorithms. variance of image between xi,j and yi,j . Meanwhile, i and j
Image Dimension Entropy show the actual position (row and column) of the pixels, and
Proposed 512 × 512 7.999 meanx and meany show the mean level of the image x and y.
Reference [47]-lena 512 × 512 7.996
The computed values for our two test images “pepper” and
“baboon” are shown in Table 8. The value for both the test
Reference [48]-lena 512 × 512 7.997
images is 1, which clearly demonstrates that the proposed
Reference [40] (Sun’s algorithm) 512 × 512 7.9965 work ensures higher security.
Reference [40] (Baptista’s algorithm) 256 × 256 7.9690
Reference [40] (Xiang’s algorithm) 256 × 256 7.9950
4.10. Maximum Difference (MD). MD is another criterion
widely adapted for image security. This test is used to find the
Table 8: SC NCC, MD, and AD computed values. actual difference that has been created between the plain and
encrypted images. The higher the value of MD depicts the
Image SC NCC MD AD
fundamental difference between the plain and encrypted
Proposed 0.8004 1 223 34.1219
images. Mathematically, maximum distance is defined as:
Tiffany 2.0954 1 255 89.1378 􏼌􏼌 􏼌􏼌
MD � MAX􏼌􏼌􏼌x(i,j) − y(i,j) 􏼌􏼌􏼌, (15)

where MAX show the actual maximum difference between

certain level of predictability arises when the IE comes out to
plain image x(i,j) at ith row and jth column and encrypted
be less than 8 bits [44, 45]. Entropy remaining close to the ideal
image y(i,j) at ith row and jth column. It is important to
value avoids IE attacks [44, 45]. Table 6 and 7 demonstrates the
achieve maximum value for a secure system. The evaluated
computed IE values for the proposed cryptosystem and its
values for our test images “Pepper” and “Tiffany” are
comparison with pre-existing schemes, respectively.
recounted in subsequent Table 8. The values of 223 for the
pepper image and 255 for the Tiffany image (Table 8) val-
4.8. Structural Content (SC). SC is involved in the provision idates the security of the proposed scheme.
of aggregate weight regarding a specific plain signal to a
coded or already present signal. A value of 1 for SC im- 4.11. Average Difference (AD). AD is another important test
plicates that the image’s quality is enhanced, while a more utilized by several research types in image processing, object
considerable value indicates that the image quality will be detection and recognition, and security application. The test
very low. Mathematically, SC is written as follows: utilizes two images, i.e., the plain image and its encrypted
image counterpart, in the formula that functions as follows:
2
􏽐M N
i�1 􏽐j�1 􏼐y(i,j) 􏼑 M N
SC � 1
2, (13) AD � 􏽘 􏽘􏼐x(i,j) − y(i,j) 􏼑, (16)
􏽐M
i�1 􏽐N
j�1 􏼐x(i,j) 􏼑 M × N i�1 j�1

where, on account of the plain and encrypted images, SC where M × N is the cumulative size of an image while
estimation is not close to unity because the encryption x(i,j) and y(i,j) are two images at the ith row and jth
14 Security and Communication Networks

column, respectively. Table 8 shows the evaluated values Conflicts of Interest

of the average difference of Pepper and Tiffany images,
which are 34.1219 and 89.1378, respectively. These results The authors declare that there are no conflicts of interest
demonstrate that a high level of security is generated regarding the publication of this article.
using the proposed system.
Authors’ Contributions
5. Ablation Study A. U., A. A. S., J. S. K., M. S., W. B., A. A., J. M., and F. A.
G. performed formal analysis and original draft preparation.
The designed image encryption algorithm is based on S. A. S. and J. A. proposed the main ideas and validated
chaotic maps. If there is a small change in key, it would analysis. A. U., J. S. K., N. R., and J.A. crystallized framework
not be possible to decrypt the original image. To re- and also revised the manuscript. All authors have read and
produce the exact results obtained in this work is to use agreed to the published version of the manuscript.
exact key. Moreover, if the size of the images is different
than it would also be not possible to get exact same References
encryption results.
[1] A. Ghaleb, F. Saeed, M. Al-Sarem et al., “Misbehavior-aware
on-demand collaborative intrusion detection system using
distributed ensemble learning for VANET,” Electronics, vol. 9,
6. Conclusion and Future Work p. 1411, 2020.
This article discusses several security parameters that can [2] M. Alkhelaiwi, W. Boulila, J. Ahmad, A. Koubaa, and
M. Driss, “An efficient approach based on privacy-preserving
be utilized to implement a robust encryption scheme and
deep learning for satellite image classification,” Remote
further perform evaluations using permutations to Sensing, vol. 13, p. 2221, 2021.
demonstrate how such methods would prevent various [3] M. Driss, D. Hasan, W. Boulila, and J. Ahmad, “Microservices
types of attacks and malicious interference. In the pro- in IoT security: current solutions, research challenges, and
posed work, we have also added multiple security layers, future directions,” Procedia Computer Science, vol. 192,
particularly multiple-dimensional chaotic iterative maps. pp. 2385–2395, 2021.
Furthermore, in the permutation process, the actual value [4] M. A. Khan, M. A. Khan Khattk, S. Latif et al., “Voting
of image pixels was changed using the 2D gingerbread classifier-based intrusion detection for iot networks,” Ad-
chaotic map to strengthen the security feature. The vances On Smart And Soft Computing, vol. 1399, pp. 313–328,
aforementioned chaotic maps generated a highly secure 2022.
ciphertext image. The proposed scheme has been tested [5] F. Masood, J. Ahmad, S. A. Shah, S. S. Jamal, and I. Hussain,
“A novel hybrid secure image encryption based on julia set of
using various security parameters, i.e., histogram anal-
fractals and 3D lorenz chaotic map,” Entropy, vol. 22, no. 3,
ysis, correlation coefficient, the mean absolute error, pp. 274–709, 2020.
differential attacks analysis, number of pixels changing [6] F. Masood, M. Driss, W. Boulila et al., “A Lightweight Chaos-
rate, unified average changing intensity, the mean square Based Medical Image Encryption Scheme Using Random
error, the peak to signal noise ratio, information entropy, Shuffling and XOR Operations,” Wireless Personal Commu-
structural content, normalized cross-correlation, maxi- nications, pp. 1–28, 2021.
mum difference, and average difference. The results [7] W. Stallings, Cryptography and Network Security, Vol. 4, E.
demonstrate the effectiveness of the proposed scheme Pearson Education India, Noida, India, 2006.
against cryptographic and differential analysis attacks. [8] H C. Cheng, Y Y. Zhi, L. GuoShiang, and H. ZengWei, “A
The scheme presented in this paper uses various light- virtual optical encryption software system for image security,”
Journal of Convergence Information Technology, vol. 6, no. 2,
weight chaos maps, and hence it could also be tested in
pp. 357–364, 2011.
real-time applications. The proposed scheme is tested in [9] H. M. Al-Najjar, “Digital image encryption algorithm based
MATLAB software; however, real-time implementation on multi-dimensional chaotic system and pixels location,”
of FPGA should also be tested. Moreover, the proposed International Journal of Computer Theory and Engineering,
scheme should be tested for audio and video applications vol. 4, no. 3, pp. 357–354, 2012.
as well. [49–52] This work is developed for symmetric key [10] A. KrBanthia and N. Tiwari, “Image encryption using pseudo
encryption only. We aim to enhance the proposed scheme random number generators,” International Journal of Com-
for asymmetric key encryption. In future, we will explore puter Application, vol. 67, no. 20, pp. 1–8, 2013.
using our suggested cryptographic techniques to secure [11] R. L. Rivest, Cryptography. In Algorithms and Complexity,
other kinds of data, such as diabetes classifications and pp. 717–755, 1990.
biological data [53]. [12] Z. Kartit, A. Azougaghe, H. K. Idrissi et al., “Applying en-
cryption algorithm for data security in cloud storage,” in
Advances in Ubiquitous Networking, pp. 141–154, Springer,
Singapore, 2016, [8].
Data Availability [13] F. Masood, W. Boulila, J. Ahmad, S. Sankar, S. Rubaiee, and
W. J. Buchanan, “A novel privacy approach of digital aerial
Data will be available upon request to the corresponding images based on mersenne twister method with DNA genetic
author. encoding and chaos,” Remote Sensing, vol. 12, p. 1893, 2020.
Security and Communication Networks 15

[14] A. Akhavan, A. Samsudin, and A. Akhshani, “Cryptanalysis of diffusion models,” Optics Communications, vol. 285,
an image encryption algorithm based on DNA encoding,” pp. 4033–4042, 2012.
Optics & Laser Technology, vol. 95, pp. 94–99, 2017. [35] Y. Wu, Y. Zhou, J. P. Noonan, and S. Agaian, “Design of image
[15] M. S. Baptista, “Cryptography with chaos,” Physics Letters A, cipher using Latin squares,” Information Sciences, vol. 264,
vol. 240, no. 1-2, pp. 50–54, 1998. pp. 317–339, 2014.
[16] R. Parvaz and M. Zarebnia, “A combination chaotic system [36] B. Norouzi, S. Mirzakuchaki, S. M. Seyedzadeh, and
and application in color image encryption,” Optics & Laser M. R. Mosavi, “A simple, sensitive and secure image en-
Technology, vol. 101, pp. 30–41, 2018. cryption algorithm based on hyper-chaotic system with only
[17] E. Solak, R. Rhouma, and S. Belghith, “Cryptanalysis of a one round diffusion process,” Multimedia Tools and Appli-
multi-chaotic systems based image cryptosystem,” Optics cations, vol. 71, no. 3, pp. 1469–1497, 2014.
Communications, vol. 283, no. 2, pp. 232–236, 2010. [37] S. M. Seyedzadeh, B. Norouzi, M. R. Mosavi, and
[18] R. Matthews, “On the derivation of a “chaotic” encryption S. Mirzakuchaki, “A novel color image encryption algorithm
algorithm,” Cryptologia, vol. 13, no. 1, pp. 29–42, 1989. based on spatial permutation and quantum chaotic map,”
[19] M. Khan and F. Masood, “A novel chaotic image encryption Nonlinear Dynamics, vol. 81, no. 1-2, pp. 511–529, 2015.
technique based on multiple discrete dynamical maps,” [38] R. Hamza and F. Titouna, “A novel sensitive image encryption
Multimedia Tools and Applications, vol. 78, Article ID 26203, algorithm based on the Zaslavsky chaotic map,” Information
2019. Security Journal: A Global Perspective, vol. 25, no. 4-6,
[20] R. A. Thiétart and B. Forgues, “Chaos theory and organiza- pp. 162–179, 2016.
tion,” Organization Science, vol. 6, pp. 19–31, 1995. [39] R. E. Boriga, A. C. Dascalescu, and A. V. Diaconu, “A new fast
[21] C. M. Reigeluth, “Chaos theory and the sciences of com- image encryption scheme based on 2D chaotic maps,” IAENG
plexity: foundations for transforming education,” in Annual International Journal of Computer Science, vol. 41, no. 4,
Meeting of the American Educational Research Association, pp. 249–258, 2014.
San Diego, CA, 2004. [40] G. Zhang and Q. Liu, “A novel image encryption method
[22] C. E. Shannon, “Communication theory of secrecy systems,” based on total shuffling scheme,” Optics Communications,
Bell system technical journal, vol. 28, pp. 656–715, 1949. vol. 284, pp. 2775–2780, 2011.
[23] S. Bone and M. Castro, A Brief History of Quantum Com- [41] J. S. Khan, W. Boulila, J. Ahmad et al., “DNA and plaintext
puting, Imperial College in London, London, UK, 1997. dependent chaotic visual selective image encryption,” IEEE
[24] F. Belkhouche and U. Qidwai, “Binary image encoding using Access, vol. 8, Article ID 159732, 2020.
one-dimensional chaotic map,” IEEE Annual Technical [42] J. S. Khan and J. Ahmad, “Chaos based efficient selective
Conference IEEE Region, vol. 5, pp. 39–43, 2003. image encryption,” Multidimensional Systems and Signal
[25] G. Chen, Y. Mao, and C. K. Chui, “A symmetric image en- Processing, vol. 30, no. 2, pp. 943–961, 2019.
cryption scheme based on 3D chaotic cat maps,” Chaos, [43] Y. Mao, G. Chen, and S. Lian, “A novel fast image encryption
Solitons & Fractals, vol. 21, no. 3, pp. 749–761, 2004. scheme based on the three-dimensional chaotic baker Map,”
[26] Y. Lu, W. Meier, and S. Vaudenay, “The conditional corre- International Journal of Bifurcation and Chaos, vol. 14,
lation attack: a practical attack on bluetooth encryption,” in pp. 3613–3624, 2004.
Annual International Cryptology Conference, pp. 97–117, [44] F. Sun, S. Liu, Z. Li, and Z. Lu, “A novel image encryption
Springer, Berlin, Heidelberg, 2005. scheme based on spatial chaos map,” Chaos, Solitons &
[27] L. Ballard, M. Green, B. De Medeiros, and F. Monrose, Fractals, vol. 38, no. 3, pp. 631–640, 2008.
Correlation-Resistant Storage via Keyword-Searchable [45] J. C. Yen and J. I. Guo, “A New chaotic key-based design for
Encryption, p. 417, IACR Cryptology ePrint Archive, 2005. image encryption and decryption,” IEEE International Sym-
[28] S. Mazloom and A. M. Eftekhari-Moghadam, “Color image posium on Circuits and Systems (ISCAS), vol. 4, pp. 49–52,
encryption based on coupled nonlinear chaotic map,” Chaos, 2000.
Solitons & Fractals, vol. 42, no. 3, pp. 1745–1754, 2009. [46] J. S. Khan, J. Ahmad, S. F. Abbasi, and S. K. Kayhan, “DNA
[29] S. Mohammad Seyedzadeh and S. Mirzakuchaki, “A fast color Sequence Based Medical Image Encryption Scheme,” in
image encryption algorithm based on coupled two-dimen- Proceedings of the 2018 10th IEEE Computer Science and
sional piecewise chaotic map,” Signal Processing, vol. 92, no. 5, Electronic Engineering (CEEC), pp. 24–29, IEEE, Colchester,
pp. 1202–1215, 2012. UK, September 2018.
[30] S. Liu, J. Sun, and Z. Xu, “An improved image encryption [47] A. Belazi, A. A. Abd El-Latif, and S. Belghith, “A novel image
algorithm based on chaotic system,” Journal of Computers, encryption scheme based on substitution-permutation net-
vol. 4, pp. 1091–1100, 2009. work and chaos,” Signal Processing, vol. 128, pp. 155–170,
[31] A. Akhshani, A. Akhavan, S. C. Lim, and Z. Hassan, “An 2016.
image encryption scheme based on quantum logistic map,” [48] X. Huang and G. Ye, “An image encryption algorithm
Communications in Nonlinear Science and Numerical Simu- based on hyper-chaos and DNA sequence,” Multimedia
lation, vol. 17, pp. 4653–4661, 2012. Tools and Applications, vol. 72, no. 1, pp. 57–70, 2014.
[32] C. Zhu and K. Sun, “Cryptanalyzing and improving a novel [49] J. Ahmad, A. Tahir, J. S. Khan, M. A. Khan, F. A. Khan, and
color image encryption algorithm using RT-enhanced chaotic Z. Habib, “A Partial Ligt-Weight Image Encryption Scheme,”
tent maps,” IEEE Access, vol. 6, Article ID 18759, 2018. in Proceedings of the 2019 IEEE UK/China Emerging Tech-
[33] A. A. Abd El-Latif, L. Li, N. Wang, Q. Han, and X. Niu, “A new nologies (UCET), pp. 1–3, Glasgow, UK, August 2019.
approach to chaotic image encryption based on quantum [50] B. A. Forouzan, Cryptography & Network Security, McGraw-
chaotic system, exploiting color spaces,” Signal Processing, Hill, New York, NY, USA, 2007.
vol. 93, pp. 2986–3000, 2013. [51] Y. Dodis, M. Stam, J. Steinberger, and T. Liu, “Indifferenti-
[34] X. Wang and L. Yang, “A novel chaotic image encryption ability of confusion-diffusion networks,” in Proceedings of the
algorithm based on water wave motion and water drop Annual International Conference on the Theory and
16 Security and Communication Networks

Applications of Cryptographic Techniques, pp. 679–704,

Springer, Berlin, Heidelberg, September 2016.
[52] J. x. Chen, Zl Zhu, C. Fu, Lb Zhang, and Y. Zhang, “An
efficient image encryption scheme using lookup table-based
confusion and diffusion,” Nonlinear Dynamics, vol. 81, no. 3,
pp. 1151–1166, 2015.
[53] J. Masood, M. Shahzad, Z. A. Khan et al., “Effective Classi-
fication Algorithms and Feature Selection for Bio-Medical
Data Using IoT,” in Proceedings of the 2020 Seventh Inter-
national Conference on Information Technology Trends (ITT),
pp. 42–47, Abu Dhabi, UAE, November 2020.