866 The International Arab Journal of Information Technology, Vol. 21, No.

5, September 2024

Image Encryption Using Modified Perturbed

Logistic Map
Subhashini Kumaran Amutha Ramachandran
Department of Electronics and Communication Engineering Department of Electronics and Communication Engineering
Sri Sai Ram Engineering College Sri Sivasubramaniya Nadar College of Engineering
India India
subhashini.ece@sairam.edu.in amuthar@ssn.edu.in

Abstract: The unique properties of chaotic maps have achieved more importance in data protection. The role of the initial
parameters in generating the chaotic sequences is vital. In the proposed approach, a new modified perturbed logistic map is
suggested. Image encryption is carried out in four steps. The chaotic sequences are attained using the proposed map at first.
Secondly, permutation on the pixels of the image is handled by a modified zig zag transformation. Thirdly, permutation effect is
further improved by dynamic quaternary DeoxyriboNucleic Acid (DNA) encoding and diffusion using DNA operations. Finally
encryption is achieved by exercising DNA decoding. It can be observed from the experiment results of statistical analysis,
differential attack analysis, Encryption Quality (EQ) analysis and performance analysis that the proposed scheme surpass
existing schemes and can resist different attacks.

Keywords: Improved zig zag transformation, image encryption, decoding, DNA operations, modified perturbed logistic map,
quaternary DNA encoding.

Received June 6, 2024; accepted August 26, 2024

https://doi.org/10.34028/iajit/21/5/8

1. Introduction which is utilized to originate the chaotic sequences.

The increase in use of multimedia images has created a The paper is aligned as: In section 2, related work is
demand for secure transmission. Multimedia has reviewed. The proposed encryption algorithm is
become a vital part of our daily lives in today’s rapidly addressed in section 3. The proposed methods’ outcome
evolving information technology landscape. Digital is examined and comparison with already existing
images have grown in popularity as a form of methods is also provided in section 4 and the results of
multimedia data due to their intuitive nature, low level the work is concluded in section 5.
of abstraction and ease of interpretation. With the arrival
of the 5G era, image utilization has increased 2. Related Works
significantly across a diverse range of industries,
Multi-media data security is a challenging task for the
including the military, medical, and transportation
traditional encryption schemes like data encryption
systems. However, as the widespread use of images
standard and advanced encryption standard to encrypt
grow, ensuring their security during transmission and
large amount of data at a reasonable speed and provide
storage has become a top priority.
security. So researchers started exploring different
To address the growing demand for image security,
encryption techniques based on chaos, transforms,
investigators have demonstrated a variety of methods,
hybrid chaotic map and Deoxyribonucleic Acid (DNA)
with image encryption emerging as the most widely
coding. Several work on chaotic map has been done
used. Chaos-based encryption schemes appear to be a
using single dimension and multi-dimension.
promising approach for safeguarding sensitive image
Ponuma and Amutha [13] performed compression
data. One-dimension chaotic map have gained
and encryption using a new 1D chaotic map formed by
popularity among the various types of chaotic systems
sine, tent and logistic map. Ponuma and Amutha [12]
due to their simple structure, implementation and low
implemented sparse coding method and compressive
computation. Numerous 1D chaotic maps have been
sensing. The authors attempted a two steps process
introduced in earlier studies, which examined how well
called scrambling and substitution using the sequences
they performed at image encryption. The sine, logistic,
generated by a 1D logistic map. Ponuma et al. [15]
sine-tangent map and hybrid maps that combine cubic
applied Discrete Wavelet Transform (DWT) and zigzag
and exponential functions have all been studied by
scan for diffusion. The chaotic sequence is used to build
researchers.
the measurement matrix. Ponuma and Amutha [14]
The contribution of this work is:
executed compression and encryption using
 A new modified perturbed logistic map is presented compressive sensing. A combination of chebyshev and
Image Encryption Using Modified Perturbed Logistic Map 867

tent map was utilized. Ye et al. [31] used DWT for chaotic system. Wei et al. [25] utilized a hyper chaotic
sparsification of plain image. The matrix was scrambled system and 2D XOR operation to diffuse the image
by row and then by column using the tent-sine map. pixels. Xu et al. [29] developed an AI based encryption
DWT and Singular Value Decomposition (SVD) are where DWT was applied on the image, followed by bit
implemented on the secret image to provide improved plane decomposition. The sequences were generated
security. Ye et al. [32] provided initial condition to using Rabinovich hyper chaotic system. Zig zag
quantum logistic map using Rivest-Shamir-Adleman transformation and then forward and reverse diffusion
(RSA) and scrambling was done by Arnold map. was performed. Jiang et al. [6] proposed a 2D map
Exclusive-OR (XOR) diffusion and modulo diffusion which improved confusion and diffusion in a random
was applied to each row and column independently. way.
Wang and Zhang [20] determined the scrambling Multi-dimensional chaotic maps are complex in
direction of image by using the multiple chaotic nature and shows some difficulty in execution and their
systems’ coordinates. Two matrices were employed in computational time is large. DNA based encryption
the diffusion step. Wang et al. [22] divided binary image techniques are being developed to overcome these
into sub images and permutation by a zigzag method. difficulties. Zhang et al. [33] achieved image encryption
One dimensional logistic self-embedding map was used by applying chaotic system and DNA coding. Wang et
for sequence generation. Xian and Wang [27] performed al. [24] manipulated the DNA sequences and chaotic
scrambling using a new fractal sorting matrix. Chen’s systems. Bitwise XOR operation was applied using
system generated two chaotic sequences for global pixel coupled map lattice. Encoding of the confused image
diffusion using XOR operation. Demirtas [3] was done using a specific DNA encode rule and
constructed a 3D bit plane array and used a Henon map. converted into a DNA array. Patel et al. [10] combined
Alawida [1], developed a new perturbed logistic map, 3D chaotic logistic map and DNA encoding. The
the chaotic range of this map is very large and its algorithm employed three keys viz. ASCII key of length
behaviour is suitable for confusion and diffusion 32 bit, a chaotic key using Chebyshev polynomial and a
process. prime key. Wang and Zhao [21] performed scrambling
One dimensional chaotic maps possess limited in blocks and DNA coding. The Chen hyper chaotic map
chaotic range which is a drawback. Many works were was used to provoke the chaotic sequence. Scrambling
proposed using hybrid chaotic maps. Khanzadi et al. [7] and then diffusion is done through different DNA
designed a random generator using logistic and tent coding rules. The cipher was achieved by the process of
chaotic maps. Wang et al. [23] demonstrated multiple DNA matrix decoding. Zhang et al. [34] combined
chaotic maps and a randomized growth approach. image signature, bit level analysis and dynamic DNA
Gayathri and Subashini [4] permuted the image pixels coding. The Hashing algorithm was used for initial
by employing circular-shift operations by fetching the value of chaotic mapping and a 3D chaotic map was
key dynamically. Liu et al. [8] created a complex used for bit level partitioning. Numerical substitution
chaotic system using Henon map and Chebyshev map. was followed by dynamic DNA coding. Wang et al. [18]
Bit plane substitution and pixel scrambling was illustrated a hybrid chaotic map with extensive range of
achieved by genetic recombination and genetic chaotic behaviour better than Logistic and Sine map.
mutation respectively. Patro et al. [11] employed the The time taken to generate the chaotic sequence was
sorted iterated sequences of the cross-coupled piecewise more. The encryption algorithm used an improved zig
linear chaotic map for diffusion. Shakiba [16] zag transformation and a quaternary DNA coding.
randomized the permutation and diffusion by a hyper Al-Hazaimeh [2] used two different chaotic maps for
chaotic map. Xu et al. [28] implemented a random walk encryption and decryption called compound chaotic
method based on hyper chaotic Lorenz system. The system. New chaotic maps that address these issues is
random walk matrix was used for image pixel needed to overcome the existing drawbacks.
scrambling and Chen’s chaotic system for diffusion Additionally, encryption algorithms depend on the
operation. Yang et al. [30] divided the input image into image pixel permutation. One of the permutation
blocks and normalized. The back propagation network methods, the standard zig zag transformation, faces few
compressed the image and then encrypted using zig zag drawbacks, including periodicity issues, could not alter
transformation and fractional order memristive system. the positions of specific pixel values in matrix as a result
Hua et al. [5] implemented an adaptive thresholding could not be used for non-square matrices. Researchers
sparsification and SWT was applied. 2D Cat map was have suggested improved zig zag transformation that
used for confusion process. Man et al. [9] executed 5D provide greater flexibility and superior permutation
Hamiltonial hyper chaotic system. This chaotic effects to overcome these limitations. DNA coding, an
sequence is the convolution kernel of CNN. Wang and innovative strategy, makes use of the similarities
Si [19] used 2D Henon map to achieve scrambling by between DNA chains and cypher sequences. However,
dynamic L shape and confusion by Arnold map. challenges remain, notably in improving the efficiency
Diffusion is applied via non-linear operation. Sheela et of DNA coding-based encryption techniques.
al. [17] performed confusion operation using 5D hyper
868 The International Arab Journal of Information Technology, Vol. 21, No. 5, September 2024

3. Proposed Encryption Algorithm used to find the control parameter using the bits 1 to 52.
52 𝑅𝑎𝑛𝑑𝑏𝑖𝑡
The proposed scheme carries out encryption by 𝜇=∑ (2)
generating the key using proposed modified perturbed 𝑡=1 2𝑡
logistic map, pixel transformation by modified zig zag The initial value is generated by utilizing the bits 53 to
method, dynamic quaternary DNA coding and 102 as follows:
operations followed by DNA decoding. 𝑅𝑎𝑛𝑑𝑏𝑖𝑡
∑102
𝑡=53 +𝜇
𝑦0 = 2𝑡 (3)
3.1. Key Generation 2

The proposed modified perturbed logistic map is given where Randbit is the value of position 𝑡. Similarly,
by the Equation (1). another four keys (128 bits) are randomly generated.
Five chaotic sequences Y1, Y2, Y3, Y4, and Y5 of size 4pq
𝜇 are generated using the same procedure. pxq is the size
𝑌𝑗+1 = 𝑎𝑏𝑠 ((𝜇 ∗ (1 − ) (1 − 𝜇2 )) 𝑚𝑜𝑑 1) (1)
𝑌𝑗 of the image. For 256x256 p=256; q=256.

where µ is the control parameter and Yj is the initial 3.2. Modified Zig Zag Transformation
value and takes values between 0 and 1. The bifurcation
The modified zig zag transformation utilizes scanning
diagram of the proposed map is shown in Figure 1. The
in two directions as shown in Figure 3. In the first
parameter µ is varied from 0 to 100. The chaotic region
process, the scanning starts from first pixel value (34 in
of the proposed map is very large when compared to
the Figure 3) of the matrix. It is continued till half the
logistic map, as a result the key space can be improved.
value of the size of the matrix. The first value is kept as
it is, the previous value is XORed with the present value
to change the pixels. The second process starts from the
last matrix value and eXclusive-NOR (XNOR)
operation is performed. The third process is shuffling
the vector 1 & 2 obtained from process one and two by
placing them in even and odd position respectively.
Three iterations are performed to permute the pixels
completely.

Figure 1. Bifurcation plot of proposed map.

Figure 3. Modified zig zag transformation.

3.3. Quaternary DNA Coding

In traditional DNA coding, the complementarity is
defined by the Watson-Crick base pairing rules, which
state the combination as adenine (a)-thymine (t) and
cytosine (c)-guanine (g). In the formation and function
Figure 2. Lyapunov plot of logistic, sine and modified perturbed of DNA the complementary base pairing rule is
logistic map. fundamental. a, g, c, t are represented as 0,1,2,3
respectively [34]. A dynamic rule is provoked to encode
The Lyapunov exponent must be positive for all
the quaternary number and decoded using another rule.
control parameter values of a chaotic map. The
Table 1 provides the 8 possible DNA coding and
proposed map gives a positive lyapunov exponent for
decoding rules [18].
control parameter values in the range 0 to 100. This
exhibits the sensitivity to initial condition. The lyapunov Table 1. DNA coding and decoding rules.
exponent of logistic and sine map is compared with the Quaternary
Rule1 Rule2 Rule3 Rule4 Rule5 Rule6 Rule7 Rule8
proposed map as shown in Figure 2. DNA
128 bits are randomly generated. Out of 128 bits 52 0 0-a 0-a 0-g 0-g 0-c 0-c 0-t 0-t
1 1-g 1-c 1-t 1-a 1-t 1-a 1-g 1-c
bits are used to generate the control parameter and 2 2-c 2-g 2-a 2-t 2-a 2-t 2-c 2-g
significand bits are only considered and Equation (2) is 3 3-t 3-t 3-c 3-c 3-g 3-g 3-a 3-a
Image Encryption Using Modified Perturbed Logistic Map 869

Step1
Chaotic sequence
generation Step2 Step2 Step3 Step4 Step5

I I1 I2 I3 I4
128 bit keys Decimal to Select rule no. Quaternary
Convert Image
Permutation Quaternary & Quaternary DNA
to Vector
Conversion DNA Encoding Permutation

µ Y0
Y1

Y2

Step6 Step7 Step8 Step9

Y3 I5 Quaternary to
Modified Perturbed I6 I7 I8
Select rule no. Decimal
Logistic Map Quaternary
DNA Chaotic & Quaternary Conversion &
DNA Diffusion
DNA Decoding represent as
matrix
Y4

Y5

Figure 4. Proposed encryption technique.

The quaternary DNA operations are performed as listed in Table 4. For example, third column of Table 4
found in the Tables 2, 3, 4, and 5. is obtained by XORing first column values with 1,
Quaternary DNA addition operation using rule 7 is 0⊕1=1 and result is represented as ‘g’ according to rule
shown in Table 2. For example in rule 7, 0⟷t; 1⟷g; 7. 1⊕1=0 and result is represented as ‘t’ according to
2⟷c; 3⟷a fourth column of Table 2 is obtained by rule 7.
adding first column values with 2 that is 0+2=2, and the Results of quaternary DNA eXclusive-NOR (DNA-
result 2 is represented as ‘c’ according to rule 7. 1+2=3 XNOR) using rule 7 is listed in Table 5. For example,
and the result 3 is represented as ‘a’ according to rule 7. fifth column of Table 5 is obtained by XNORing first
column values with 3, 0⨀3 and result is represented as
Table 2. Quaternary DNA addition using rule 7. ‘t’ according to rule 7. 1⨀3=1 and result is represented
Add 0-t 1-g 2-c 3-a as ‘g’ according to rule 7.
0-t t g c a
1-g g c a t Table 5. Quaternary DNA-XNOR using rule 7.
2-c c a t g
3-a a t g c XNOR 0-t 1-g 2-c 3-a
0-t a c g t
1-g c a t g
Results of quaternary DNA subtraction using rule 7 2-c g t a c
is listed in Table 3. For example, second column of 3-a t g c a
Table 3 is obtained by subtracting 0 from first column
values, 0-0=0 and result is represented as ‘t’ according 3.4. Algorithm
to rule 7. 1-0=1 and result is represented as ‘g’
The block diagram of the image encryption is shown in
according to rule 7.
Figure 4.
Table 3. Quaternary DNA subtraction using rule7.
 Step 1. Chaotic sequence generation. Generate five
Sub
0-t
0-t
t
1-g
a
2-c
c
3-a
g
chaotic sequences of size 4pxq using the proposed
1-g g t a c modified perturbed logistic map, where pxq is the
2-c c g t a size of the image.
3-a a c g t
 Step 2. Permutation and convert image to vector.
Table 4. Quaternary DNA XOR using rule 7. Apply modified zig zag transformation to permute
XOR 0-t 1-g 2-c 3-a the pixels of image I to get I1. The resultant matrix is
0-t t g c a arranged as a vector I2.
1-g g t a c  Step 3. Decimal to quaternary conversion. Convert
2-c c a t g
3-a a c g t the pixel values in I2 into its equivalent quaternary
representation to get I3. The vector dimension of I3 is
Results of quaternary DNA-XOR using rule 7 is 4pq.
870 The International Arab Journal of Information Technology, Vol. 21, No. 5, September 2024

Conversion of decimal to quaternary. g, a, t, c, c, t, t, c, t] is obtained and the decoding rule

Consider the permuted pixels of I2 vector as [12, 5, 8, vector DNAdecodingrule(j)= [2, 4, 6, 7, 1, 3, 5, 8, 2, 4] is
22]. The decimal values are converted into quaternary generated using Equation (9). The decoded vector is
(base 4) as shown in Figure 5. I7(j)=[a, a, g, t, c, a, c, t, c, g]. Convert I7(j) into its
quaternary value [3, 3, 1, 0, 2, 3, 2, 0, 2, 1].
 Step 9. Quaternary to decimal conversion and
represent as matrix. Convert vector 𝐼7 into its
equivalent decimal value as I8. Group I7(j) into a
group of 4 values. For example [3, 3, 1, 0, 2, 3, 2, 0,
2, 1] is divided into [3, 3, 1, 0], [2, 3, 2, 0]. Convert
each group into its equivalent decimal value to
generate the vector I8(j)=[244, 184, …]. Construct
vector I8 as a matrix which results in the cipher
image.

4. Simulation Results and Discussions

Figure 5. Decimal to quaternary conversion process. Simulation results are obtained using images from
University of Southern California-Signal and Image
Quaternary equivalent of 12, 5, 8 and 22 is 30, 11, 20 Processing Institute (USC-SIPI) database. Grayscale
and 112 respectively. The quaternary values are images of size 256x256, 512x512 and 1024x1024 are
represented as [0030, 0011, 0020, 0112], using 4 digits. used. The images used in the simulations are Clock,
Concatenation of these values gives the vector I3 as Peppers, Baboon and Man. Image encryption is done
[0 0 3 0 0 0 1 1 0 0 2 0 0 1 1 2]. using three maps, namely proposed new modified
 Step 4. Select rule no. and quaternary DNA encoding. perturbed logistic map, Hybrid Logistic Sine
Select anyone of the eight DNA encoding rule Exponential (HLSE) map [18] and perturbed logistic
dynamically to encode vector 𝐼3 using the following map [1]. The performance of the image encryption
equation to get I4 of dimension 4pq. algorithm is verified using security analysis,
encryption analysis, statistical test and differential
𝐷𝑁𝐴𝑒𝑛𝑐𝑜𝑑𝑖𝑛𝑔𝑟𝑢𝑙𝑒 (𝑗) = (4)
𝑓𝑙𝑜𝑜𝑟(8 ∗ 𝑌1 (𝑗)) + 1, 𝑗 = 1,2, … ,4𝑝𝑞
attack analysis. The encryption algorithm is executed
using a system with AMD 5600H CPU @3.3GHz
 Step 5. Quaternary DNA permutation. Sort Y2 in and 8 GB RAM, PyCharm Integrated Development
increasing order and procure the index number which Environment (IDE) and MATLAB R2022a. The
is used to permute I4 and get vector I5 of size 4pq. input and the respective cipher obtained by applying
𝐼5 (𝑗) = 𝐼4 (𝑆𝑜𝑟𝑡𝑒𝑑𝑖𝑛𝑑𝑒𝑥 (𝑗)), 𝑗 = 1,2, … , 4𝑝𝑞 (5) the proposed algorithm are shown in Figure 6.

 Step 6. DNA Chaotic. Convert Y3 into Y3DNAchaotic

using the following equations,
𝑖𝑓 0 ≤ 𝑌3 < 0.25, 𝑡ℎ𝑒𝑛 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 = 0
𝑖𝑓 0.25 ≤ 𝑌3 < 0.5, 𝑡ℎ𝑒𝑛 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 = 1 (6)
}
𝑖𝑓 0.5 ≤ 𝑌3 < 0.75, 𝑡ℎ𝑒𝑛 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 = 2
a) Plain image of b) Cipher image of c) Plain image of d) Cipher image of
𝑖𝑓 0.75 ≤ 𝑌3 < 1, 𝑡ℎ𝑒𝑛 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 = 3 Clock. Clock. Man. Man.

 Step 7. Quaternary DNA diffusion. Select one of the

four DNA functions dynamically using the following
rule to perform diffusion and obtain vector I6.
𝐷𝑁𝐴𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑟𝑢𝑙𝑒 (𝑗) = 𝑓𝑙𝑜𝑜𝑟(4 ∗ 𝑌4 (𝑗)) + 1, (7)
𝑖𝑓 𝐷𝑁𝐴𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑟𝑢𝑙𝑒 (𝑗) = 1, 𝑡ℎ𝑒𝑛 𝐴𝑑𝑑(𝐼5(𝑗), 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 (𝑗)) e) Plain image of f) Cipher image of g) Plain image of h) Cipher image of
Peppers. Peppers. Baboon. Baboon.
𝑖𝑓 𝐷𝑁𝐴𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑟𝑢𝑙𝑒 (𝑗) = 2, 𝑡ℎ𝑒𝑛 𝑆𝑢𝑏(𝐼5(𝑗), 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 (𝑗))
= 𝐼6(𝑗)
𝑖𝑓 𝐷𝑁𝐴𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑟𝑢𝑙𝑒 (𝑗) = 3, 𝑡ℎ𝑒𝑛 𝑋𝑂𝑅(𝐼5(𝑗), 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 (𝑗)) (8) Figure 6. Plain, Cipher images for each one of these: Clock, Man,
𝑖𝑓 𝐷𝑁𝐴𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑜𝑛𝑟𝑢𝑙𝑒 (𝑗) = 4, 𝑡ℎ𝑒𝑛 𝑋𝑁𝑂𝑅(𝐼5(𝑗), 𝑌3𝐷𝑁𝐴𝑐ℎ𝑎𝑜𝑡𝑖𝑐 (𝑗))} Peppers and Baboon.
 Step 8. Select rule no. and quaternary DNA decoding:
Generate the DNA decoding rule using Equation (9) 4.1. Statistical Analysis
to decode vector I6 to I7 Statistical attacks happen due to the correlation that
𝐷𝑁𝐴𝑑𝑒𝑐𝑜𝑑𝑖𝑛𝑔𝑟𝑢𝑙𝑒 (𝑗) = exists among the neighboring pixels. The proposed
(9) technique is assessed using statistical analysis like
𝑓𝑙𝑜𝑜𝑟(8 ∗ 𝑌5 (𝑗)) + 1, 𝑗 = 1,2, … , 4𝑝𝑞
histogram, correlation, entropy analysis, chi-square test,
For example, after diffusion operation the vector I6=[t, National Institute of Standards and Technology (NIST)
Image Encryption Using Modified Perturbed Logistic Map 871

test and Peak Signal to Noise Ratio (PSNR) and Mean of correlation coefficient of the proposed method in
Square Error (MSE) analysis in the next section. horizontal and vertical direction is ‘-0.018’ and ‘-
0.0109’ respectively, which is lesser compared to the
4.1.1. Histogram Analysis results of [16, 18, 28, 32]. Figure 8-a), (b) and (c) depicts
Histogram illustrates the pixel frequency distribution. the pixel correlation plot of peppers image in three
The histogram of plain and cipher is presented in Figure directions. The pixels are grouped in the diagonal
7. direction and infers very strong correlation. The scatter
It is observed that there exists an uneven distribution plot of the cipher image displayed in Figure 8-d), (e) and
of image pixels in the plain histogram and it is prone to (f) is sporadic and combat attacks based on correlation.
statistical attacks. The encrypted image exhibit even Table 6. Correlation analysis of Peppers image
distribution for all test images and hence the proposed Algorithm Horizontal Vertical Diagonal
technique can withstand such attacks. Moreover, the Plain image 0.9721 0.9767 0.9651
snooper could not pry input from cipher. [18] 0.0003 0.0003 -0.0034
[1] -0.0064 -0.0003 0.0061
Proposed -0.018 -0.0109 0.0013
[10] -0.0022 -0.0016 0.0046
[32] 0.0139 0.0054 0.0153
[11] - 0.0013 0.0019 0.0009
[16] 0.002553 −0.0138 −0.0028
[28] 0.0084 0.0045 −0.0023
a) Plain image of b) Histogram image of c) Cipher image of d) Cipher histogram
Clock. Clock. Clock. image of Clock.

a) Horizontal direction of the b) Vertical direction of the c) Diagonal direction of the

e) Plain image of f) Histogram image of g) Cipher image of h) Cipher histogram test Peppers image. test Peppers image. test Peppers image.
Peppers. Peppers. Peppers. image of Peppers.

i) Plain image of j) Histogram image of k) Cipher image of l) Cipher histogram d) Horizontal direction of the e) Vertical direction of the f) Diagonal direction of the
Baboon. Baboon. Baboon. image of Baboon. cipher Peppers image. cipher Peppers image. cipher Peppers image.

Figure 7. Plain, histogram, cipher, cipher histogram images for each Figure 8. Pixel correlation plot of test and cipher for image Peppers
one of these: Clock, Peppers, and Baboon. in a horizontal, vertical, diagonal directions.

4.1.2. Correlation Analysis 4.1.3. Global Entropy Analysis

The correlation between neighbouring pixels of an Information entropy, used to find the uncertainty present
image can be found using the following equations: in cipher image. Entropy value of a cipher is ideally
equal to 8. The entropy of the test and cipher image are
𝑐𝑜𝑣(𝑥, 𝑦)
𝑅𝑥𝑦 = (10) calculated using the following equation:
√𝐴(𝑥)𝐴(𝑦)
𝐿
1
where, 𝐻(𝑧) = ∑ 𝑝(𝐺𝑉𝑖 )𝑙𝑜𝑔2 (13)
𝑁 𝑝(𝐺𝑉𝑖 )
𝑖=0
1
𝑐𝑜𝑣(𝑥, 𝑦) = ∑(𝑥𝑗 − 𝐸(𝑥))(𝑦𝑗 − 𝐸(𝑦)) (11)
𝑁
𝑗=1
p(GVi)is the probability of the image pixels and 𝐿 is the
𝑁
maximum pixel value.
1 (12)
𝐴(𝑦) = ∑(𝑦𝑗 − 𝐸(𝑦))2 Table 7. Global entropy.
𝑁
𝑗=1
Cipher image
Test image Plain image
The correlation coefficient is measured in horizontal, [18] [1] Proposed
Clock 256x256 6.7057 7.9971 7.9973 7.9973
vertical and diagonal direction by arbitrarily selecting Peppers 512x512 7.5937 7.9992 7.9991 7.9993
10000 pixels. Table 6 provides the correlation values of Man 1024x1024 7.5237 7.9998 7.9998 7.9998
the peppers image for the proposed and existing
methods. The values are approximately equal to 1 for It is inferred from Table 7 that the cipher image
plain image and ascertain high correlation. The entropy values for test images of different dimension are
correlation values of the cipher obtained for the very close to 8. The average entropy 7.9989 of the
proposed algorithm are less than ‘0.01’ indicating that proposed approach is close to ideal value.
the correlation is minimum and close to ‘0’. The value
872 The International Arab Journal of Information Technology, Vol. 21, No. 5, September 2024

Table 8. Comparison of global entropy of Baboon image. critical value. This proves that the histogram of the
Test Plain Cipher image cipher holds uniform distribution and the algorithm
image image [18] [1] Proposed [23] [24] [33] works well with all three maps.
Baboon
7.3583 7.9927 7.9993 7.9993 7.9974 7.9969 7.9992
512x512
Table 10. Chi-square test.
Test image [18] [1] Proposed
The entropy value of the Baboon image of the Clock 256x256 261.0859 248.6172 242.5156
proposed algorithm is compared with other algorithms Baboon 512x512 254.6895 253.5898 271.4922
in the literature and is listed in Table 8. The proposed Man 1024x1024 245.481 286.3892 282.6855
approach gives 7.9993, a high entropy value compared
to existing methods. Chi-square value of the peppers image for the
Local entropy is used to test the randomness of proposed algorithm and other algorithms in the
encrypted image. The global Shannon entropy measure literature are shown in Table 11. The Chi-square value
is inconsistent for images of varying dimension and the of the peppers image using the proposed method is
local entropy measure provides a relatively fair 227.32 which is 19% lesser than 264.77 as reported in
comparison for image randomness among multiple [10]. The chi-square value of the proposed method is
images. lesser than [4, 11, 20].
𝑘 Table 11. Comparison of chi-square value of Pepper image.
𝐻(𝐵𝑖 )
𝐻𝑘,𝑇𝐵 (𝐵) = ∑ (14) Image [18] [1] Proposed [10] [20] [11] [4]
𝑘
𝑖=1 Peppers 275.07 310.67 227.32 264.77 287.21 271.9 244.49
where k is the number of non-overlapping blocks and
H(Bi) is the Shannon entropy. The image is divided into 4.1.5. NIST Test
𝑘 blocks and each block consist of TB pixels. For k=30 NIST is computational series of 15 test that validates the
and TB=1936 the local entropy is calculated. The randomness of the cipher image by using their binary
average entropy of all blocks gives the local entropy input to test for randomness and to determine its p-
[26]. value. It determines whether the sequence is predictable,
Table 9. Local entropy analysis.
if the value is less than or equal to 0.01 it fails the test.
NIST-Special Publication (NIST-SP) 800-22 revision
Test image [18] [1] Proposed
Clock 256x256 7.8998 7.9059 7.9025
1a test suite is used to carry out the test. Table 12 gives
Baboon 512x512 7.9041 7.903 7.9042 the 15 NIST test of the proposed modified perturbed
Peppers 512x512 7.9021 7.9033 7.9035 logistic method for the clock cipher image. The test
Man 1024x1024 7.8806 7.9046 7.9021
results show that the proposed method passes all the
The test results in Table 9 shows that the proposed NIST test and has p-value>0.01.
method is equal to the ideal value (7.902469317) and Table 12. NIST test results for Clock cipher image.
passes the test for the images clock, baboon and peppers Proposed
with a significance level α=0.05. Test
P-value Result
Frequency 0.324931 Pass
Block frequency 0.199739 Pass
4.1.4. Chi-Square Test Runs 0.044259 Pass
This test provides the pixel uniformness of the cipher Longest 0.146922 Pass
Rank 0.962089 Pass
image. The chi-square value is calculated by the FFT 0.019060 Pass
equation, Non-overlapping template 0.809319 Pass
Overlapping template 0.631976 Pass
255 (𝑜𝑏 − 𝑒𝑥)2 Linear complexity 0.281564 Pass
𝑥2 = ∑ (15)
𝑖=0 𝑒𝑥 Serial test P-value 1 0.525440 Pass
Serial test P-value 2 0.109853 Pass
The observed value ‘ob’ is the count of each gray value Approximate entropy 0.656514 Pass
Cumulative sums-forward 0.604405 Pass
has occurred in the cipher image. The expected value Cumulative sums-reverse 0.529498 Pass
‘ex’ is given by, Random excursions test (X=1) 0.443414 Pass
𝑝𝑥𝑞 Random excursions test (X=-1) 0.340473 Pass
𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑣𝑎𝑙𝑢𝑒, 𝑒𝑥 = (16) Random excursions variant test (X=1) 0.425030 Pass
256 Random excursions variant test (X=-1) 0.732439 Pass
where p and q is the value of available rows and columns
in the image and expected value of a 256x256 image is 4.1.6. PSNR and MSE Analysis
256. MSE and PSNR are calculated between the plain and the
The critical chi-square value is 𝜘2=293.2478 for 255 cipher image to check for any relation between them.
degrees of freedom at 5% significance level or α=0.05 MSE finds the average squared pixel difference
[10]. For a gray scale image the degrees of freedom is corresponding to plain and cipher, higher the value the
related to the pixel intensity values. The test results greater is the encryption effect. If the PSNR is low,
indicate in Table 10 that all cipher are less than the superior is the Encryption Quality (EQ).
Image Encryption Using Modified Perturbed Logistic Map 873

MSE is defined as: PSNR is less than 10dB, it is not possible to retrieve the
1 𝑝 𝑞 plain image from the cipher.
𝑀𝑆𝐸 = ∑ ∑ (𝐶1 (𝑘, 𝑙) − 𝐶2 (𝑘, 𝑙))2 (17)
𝑝𝑥𝑞 𝑘=1 𝑙=1
4.2. Differential Attack Analysis
C1(k, l) and C2(k, l) are the individual pixels of plain and
cipher respectively. The encryption algorithm must yield a totally different
PSNR is defined as: encrypted image even there is only one pixel deviation
in the plain image. The sensitivity of the algorithm can
2552
𝑃𝑆𝑁𝑅 = 10 ∗ 𝑙𝑜𝑔10 ( ) (18) be proved by finding Number of Pixel Change Rate
𝑀𝑆𝐸
(NPCR) and Unified Average Change Intensity (UACI)
where 255 is the maximum pixel value. values using the Equations (19) and (21) respectively.
Table 13. MSE values. 1
𝑁𝑃𝐶𝑅 = ∑ 𝐷(𝑗, 𝑘) ∗ 100% (19)
Test image [18] [1] Proposed 𝑝𝑥𝑞
𝑗,𝑘
Clock (256x256) 12073 12111.5 12101.67
Baboon (512x512) 7235 7276.58 7240.42 0, 𝐼1 (𝑗, 𝑘) = 𝐼2 (𝑗, 𝑘)
where 𝐷(𝑗, 𝑘) = { (20)
Man (1924x1024) 10312.99 10298.62 10303.05 1, 𝐼1 (𝑗, 𝑘) ≠ 𝐼2 (𝑗, 𝑘)

The MSE is computed for different test images as The unified average change intensity is calculated using
shown in Table 13. The mean square value of the the equation,
proposed approach is higher than the HLSE method and 1 ∑𝑗,𝑘|𝐼1 (𝑗, 𝑘) − 𝐼2 (𝑗, 𝑘)|
𝑈𝐴𝐶𝐼 = ∗ 100% (21)
lesser than the perturbed logistic method. This indicates 𝑝𝑥𝑞 𝑚𝑎𝑥
that the proposed map produces equivalently large pixel where I1 and I2 are the two ciphers of the plain images
differences between plain and cipher as HLSE map and having one pixel difference and ‘max’ stands for
perturbed logistic map. maximum intensity of the pixel. The one-pixel
Table 14. Comparison of MSE of Peppers image. difference in plain image is achieved by arbitrarily
Image [18] [1] Proposed [4] [16] [11]
selecting a pixel value and adding one to it.
Peppers 8408.1 8400.34 8389.80 8260 5413.9 8465.8
Table 17. NPCR values.
The MSE value of the peppers image for the Test image [18] [1] Proposed
Clock 256x256 99.5697 99.5926 99.6338
proposed modified perturbed logistic method is Baboon 512x512 99.6273 99.6201 99.6178
compared with the other algorithm and is shown in Man 1024x1024 99.6126 99.6070 99.6069
Table 14. It is inferred that the MSE values of the
proposed method is slightly less than HLSE and Test results in Table 17 renders the NPCR values for
perturbed logistic method. The MSE value of the images of different dimension and the proposed results
proposed algorithm for peppers image is 1.56% and are above critical value. The average NPCR of the
43.12% higher than [4, 16] respectively. This shows EQ proposed algorithm is 99.6148, which is higher than [1,
is high. 18]. The sensitivity of the proposed encryption
algorithm to one pixel variation is proved. This
Table 15. PSNR values. emphasizes the protection against differential attacks.
Test image [18] [1] Proposed
Clock 256x256 7.3125 7.30 7.30 Table 18. Comparison of NPCR values of Peppers image.
Baboon 512x512 8.8838 9.51 9.53
Test image [18] [1] Proposed [32] [16] [20]
Man 1024x1024 9.5364 8.00 8.00
Peppers 99.6075 99.6307 99.6201 99.6086 99.6067 99.6108

Table 15 indicates that the PSNR of the proposed Test results in Table 18 shows NPCR of proposed
approach is approximately same as that of perturbed algorithm for peppers image is less than perturbed
logistic method and 2.614% less than the HLSE method. logistic method and higher than all other methods. For
The result also illustrates that for different image image dimension 256x256, 512x512 and 1024x1024,
dimension (256x256, 512x512 and 1024x1024) PSNR the critical NPCR is 99.5693%, 99.5893% and
is less than 10dB for all three maps, indicating high EQ. 99.5994% respectively. The proposed approach can
Table 16. Comparison of PSNR of Peppers image. withstand the differential attack as the NPCR is above
Test image [18] [1] Proposed [4] [16] [11]
the critical value for all test images.
Peppers 8.88 8.89 8.89 9.0234 10.7957 8.8541
Table 19. UACI values.
Test results in Table 16 shows the PSNR value for Test image [18] [1] Proposed
Clock 256x256 33.5536 33.4741 33.5384
peppers image. The PSNR of the modified perturbed Baboon 512x512 33.5232 33.4265 33.5468
logistic method is 1.49%, 19.36% less than [4, 16]. MSE Man 1024x1024 33.4592 33.4624 33.4806
and PSNR are contrarily equivalent to each other. An
increase in MSE value decreases the PSNR. As the
874 The International Arab Journal of Information Technology, Vol. 21, No. 5, September 2024

255 𝑔𝑖
Test results in Table 19 reveals that the proposed 𝐸𝑛𝑐𝑟𝑦𝑝𝑡𝑖𝑜𝑛 𝑄𝑢𝑎𝑙𝑖𝑡𝑦 𝐸𝑄 = ∑
256
(23)
algorithm gives highest UACI value for different plain 𝑖=0

text. The average value of the proposed method 33.5125 Table 23. EQ values for different test images.
is also better than other two methods. This ensures that
Test image [32] [11] Proposed
the modified perturbed logistic method produces a Clock 256x256 242.9063 242.9922 240.6484
totally altered cipher even for a one pixel variation. Baboon 512x512 774.90 775.84 778.1641
Man 1024x1024 2328.5 2329.7 2333.8
Table 20. Comparison of UACI values of Peppers image.
Test image [18] [1] Proposed [32] [16] [20] Test results in Table 23 unveils that the EQ values of
Peppers 33.4526 33.4508 33.5203 33.4398 33.4332 33.5173 proposed computation method is greater than HLSE
method and perturbed logistic method for Baboon and
The UACI of peppers image is compared with [16, Man images. High EQ values confirm the EQ of the
20, 32] in Table 20 and the proposed method shows cipher.
highest value. The critical UACI value can lie in the
range (33.2824%, 33.6447%), (33.3730%, 33.5541%) Table 24. Comparison of EQ for Peppers image.
and (33.4183%, 33.5088%) for image dimension Test image [18] [1] Proposed
256x256, 512x512 and 1024x1024 respectively. Peppers 568 565.96 570.9531

It is inferred from Table 24 that the proposed

4.3. Encryption Quality Analysis
modified perturbed logistic method gives higher value
Maximum Deviation (MD) is a metric that is use find of EQ compared with other algorithms.
the deviation of density of pixel values between cipher
and plain image. If the MD value is high, it implies that 4.3.3. Irregular Deviation
there is high deviation between the pixel density of The Irregular Deviation (ID) is a metric used to measure
cipher and plain image. EQ is the metric which the irregularity or non-uniformity of pixel distribution
evaluates the pixel intensity deviation between the in a cipher image histogram. The ID value is used to
original and cipher. evaluate the evenness of pixel distribution in the cipher
image. A lower ID value indicates a more uniform or
4.3.1. Maximum Deviation
even distribution of pixels in the image, whereas a
MD can be defined as: higher ID value indicates a more irregular or non-
𝑔0 + 𝑔255 254 uniform distribution.
𝑀𝑎𝑥 𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 = + ∑ 𝑔𝑖 (22)
2 255
𝑖=1
𝐼𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑟𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 = ∑|𝑔𝑖 − 𝛾ℎ | (24)
where gi is given as |Ii-Ci| where Ii and Ci are each pixel 𝑖=0
value frequency in plain and cipher image, g0 and g255
stands for number of pixels having 0 and 255 as where 𝛾ℎ is the histogram average.
grayscale value respectively. Table 25. Comparison of ID.
Table 21. Comparison of MD values. Test image Proposed [18] [1]
Clock 256x256 32881 32802 32807
Test image [18] [1] Proposed
Clock 256x256 61678 61701 61087
Peppers 512x512 143232.5 142866 144127.5 The results infer that the proposed method produced
Man 1024x1024 576981.5 577334.5 578517 equivalently comparable result as that of existing
methods as shown in Table 25.
MD value of proposed algorithm is compared in
Table 21 with HLSE method and perturbed logistic 4.3.4. Deviation from Uniform Histogram
method.
The deviation from uniform histogram measures the
Table 22. Comparison of MD for Baboon image. discrepancy between the cipher histogram and an ideal
Test image [18] [1] Proposed [20] uniform histogram.
Baboon 196214 196590 197185 46540
1
𝐼𝑚𝑎𝑔𝑒 ℎ𝑖𝑠𝑡𝑜𝑔𝑟𝑎𝑚 𝐼ℎ = {256 𝑝𝑥𝑞, 0 ≤ 𝑖 ≤ 255 (25)
Test results in Table 22 indicates that the proposed 0, 𝑒𝑙𝑠𝑒𝑤ℎ𝑒𝑟𝑒
modified perturbed logistic map gives high MD value
∑255
ℎ𝑖=0 |𝐼ℎ𝑖 − 𝐼ℎ |
for Baboon image compared to other methods. The MD 𝐷𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛ℎ𝑖𝑠𝑡𝑜𝑔𝑟𝑎𝑚 = (26)
of Baboon image calculated using the proposed method 𝑝𝑥𝑞
is 123% higher than that of [20]. where 𝐼ℎ𝑖 is the cipher image gray scale value at position
i. The Deviationhistogram should be less to indicate that the
4.3.2. Encryption Quality (EQ) distribution is uniform.
EQ is represented by, The results infer a minimum deviation and thereby
proves that the pixels are uniformly distributed in the
Image Encryption Using Modified Perturbed Logistic Map 875

proposed method as indicated in Table 26. image CI2 provides insights into the impact of key
interchange on decryption outcomes. This shows that
Table 26. Comparison of DUH.
the algorithm is highly sensitive to keys.
Test image Proposed [18] [1]
Clock 256x256 0.04965 0.05035 0.04925
Peppers 512x512 0.0240 0.0301 0.0276
Baboon 512x512 0.0252 0.0299 0.0247
Man 1024x1024 0.0130 0.0131 0.0131

4.3.5. Contrast Analysis c) Absolute

a) Cipher CI1 b) Cipher CI2 difference d) Decrypt CI1 e) Decrypt CI2
using Key1. using Key2. between CI1 and using Key2. using Key1.
In contrast analysis the variation in gray level is used to CI2.
find patterns. The analysis uses co-occurrence matrix,
Figure 9. Key sensitivity analysis using two different Keys.
CoOccurrence (i, j) to identify specific patterns in the
image.
4.4. Performance Analysis
256 256

𝐶𝑜𝑛𝑡𝑟𝑎𝑠𝑡 = ∑ ∑|𝑖 − 𝑗|2 𝐶𝑜𝑂𝑐𝑐𝑢𝑟𝑟𝑒𝑛𝑐𝑒(𝑖, 𝑗) (27) 4.4.1. Permutation Effect Analysis

𝑖=0 𝑗=0
The permutation effect is shown for the modified zig
where i and j represent 8 bit gray levels. zag transformation of the image matrix and for a total of
three iterations. Figure 10 depicts that the number of
Table 27. Contrast analysis for different test images.
iterations increases the permutation effect. The
Test image [18] [1] Proposed transformation is more uniform and outperforms the
Clock 256x256 10910.469 10855.3661 10876.5668
Peppers 512x512 10863.654 10842.007 10861.8307 pervious iterations.
Baboon 512x512 10882.011 10890.3316 10846.965
Man 1024x1024 10919.678 10887.0349 10908.0449

The contrast analysis of proposed method shown in

Table 27 infers comparable performance with that of
existing techniques.
b) Modified zig zag c) Modified zig zag d) Modified zig zag
a) Plain image of
4.3.6. Key Space Analysis Clock.
transformation output transformation output transformation output
for iteration 1. for iteration 2. for iteration 3.
The key space of the proposed method for five 128 bit
keys is 2128 x2128 x2128 x2128 x2128 which is 2640 >2100 and
sufficient to resist brute force attack. The larger key
space increases ability of the algorithm to prevent
intensive attacks.
f) Completely g) Completely h) Completely
Table 28. Key space analysis. e) Plain image of
encrypted image for encrypted image for encrypted image for
Clock.
[18] [1] Proposed iteration 1. iteration 2. iteration 3.
10224 2128 2640 Figure 10. Permutation effect of modified zig zag transformation and
completely encrypted output for different iterations.
Test results in Table 28 compares the key space of
the proposed modified perturbed logistic method and 4.4.2. Correlation Effect of Permutation
other methods in the literature.
The correlation coefficients of plain image after zig zag
4.3.7. Key Sensitivity Analysis transformation is tabulated in Table 29. These values are
given for iterations one, two and three.
Key sensitivity is a phenomenon where even slight
modifications in encryption or decryption keys results Table 29. Correlation values of permuted Peppers image.
in significant changes in the resulting ciphertext or Iteration Direction Plain image Permuted image
Horizontal 0.9721 -0.00148
plaintext. Key sensitivity analysis involves generating Iteration=1 Vertical 0.9767 -0.00139
two distinct keys: Diagonal 0.9651 0.00061
Horizontal 0.00059
Key1=504b444ea02e166342b1e53a5aad76f87b0cfda Iteration=2 Vertical 0.00075
Key2=504b444ea02e166342b1e53a5aad76f87b0ceda. Diagonal 0.00097
Horizontal 0.00032
Key1 and Key2 are used to encrypt the original image, Iteration=3 Vertical 0.00038
Diagonal -0.00347
yielding two distinct cipher images, CI1 and CI2 as
shown in Figure 9. To quantify the visual dissimilarity
The correlation values decrease as the iteration
caused by key divergence, the absolute difference
number increase.
between CI1 and CI2 is computed. Moreover, employing
Key2 to decrypt cipher image CI1 and Key1 for cipher
876 The International Arab Journal of Information Technology, Vol. 21, No. 5, September 2024

4.5. Robustness Analysis 4.5.3. Speed Analysis

4.5.1. Noise Attack Analysis A faster algorithm is needed to handle large data like
images quickly and efficiently. Adding more rounds to
The cipher images underwent a deliberate injection of a cipher increases security but slows performance. On
salt and pepper noise at varying intensities: 0.01, 0.05, the other hand, reducing rounds improves performance
0.1, and 0.15. Following this, the noisy cipher images but lowers security.
were subjected to the decryption process and are
presented in Figure 11. Decrypted images retain Table 30. Speed analysis.
noticeable traces of the introduced noise. However, it is Test image [18] [1] Proposed
worth highlighting that despite the noise, the images 256x256 4.7973 4.5112 4.8562
512x512 9.1884 12.0848 11.6921
remain quite recognizable. This significant observation 1024x1024 32.2786 42.1770 43.1967
underscores the algorithm’s capacity to effectively
withstand noise interference, showcasing its robust anti- Table 30 infers that the time taken in seconds by the
noise property. proposed method to execute the program is comparable
with the other existing methods.

4. Conclusions
In this paper, an image encryption algorithm using
a) Encrypted image b) Encrypted image c) Encrypted image d) Encrypted image modified perturbed logistic map is proposed. The
with noise intensity with noise intensity with noise intensity with noise intensity
0.01. 0.05. 0.1. 0.15. performance of the proposed algorithm is evaluated
using different performance metrics like EQ, MD, chi
square test, NPCR, UACI, correlation analysis, local
entropy, global entropy, noise attack analysis, cropping
attack analysis, key space and key sensitivity. The
e) Decrypted image f) Decrypted image g) Decrypted image h) Decrypted image
metric values obtained using proposed method is
with noise intensity with noise intensity with noise intensity with noise intensity compared with other encryption schemes available in
0.01. 0.05. 0.1. 0.15.
the literature. The key space of the proposed scheme is
Figure 11. Noise attack analysis of encrypted and decrypted Clock large. Hence the proposed work can resist brute force
image using different noise intensity levels.
attack.
4.5.2. Cropping Attack Analysis
References
The original clock image is introduced with varying
degrees of occlusion (1/32, 1/16, 1/8, and 1/4) as shown [1] Alawida M., “A Novel Chaos-Based Permutation
in Figure 12-a), (b), (c), and (d). The subsequent step for Image Encryption,” Journal of King Saud
involved decrypting the resulting cropped cipher University-Computer and Information Sciences,
images, which are depicted in Figure 12-e), (f), (g), and vol. 35, no. 6, pp. 101595, 2023.
(h). Despite the intentional occlusion, the decrypted https://doi.org/10.1016/j.jksuci.2023.101595
images displayed remarkable clarity, showcasing the [2] Al-Hazaimeh O., Al-Jamal M., Bawaneh M.,
algorithm’s resilience against cropping attacks. This Alhindawi N., and Hamdoni B., “A New Image
shows the algorithm’s ability to maintain image fidelity, Encryption Scheme Using Dual Chaotic Map
even in scenarios involving significant content Synchronization,” The International Arab Journal
concealment. of Information Technology, vol. 18 no. 1, pp. 95-
102, 2021.
https://www.iajit.org/portal/PDF/Vol%2018,%20
No.%201/19572.pdf
[3] Demirtas M., “A Novel Multiple Grayscale Image
Encryption Method Based on 3D Bit-Scrambling
and Diffusion,” Optik, vol. 266, pp. 169624, 2022.
a) Occulted cipher b) Occulted cipher c) Occulted cipher d) Occulted cipher
with proportion 1/32. with proportion 1/16. with proportion 1/8. with proportion 1/4. https://doi.org/10.1016/j.ijleo.2022.169624
[4] Gayathri J. and Subashini S., “A Spatiotemporal
Chaotic Image Encryption Scheme Based on Self
Adaptive Model and Dynamic Keystream
Fetching Technique,” Multimedia Tools and
Applications, vol. 77, no. 19, pp. 24751-24787,
e) Decrypted image f) Decrypted image g) Decrypted image h) Decrypted image
with proportion 1/32. with proportion 1/16. with proportion 1/8. with proportion 1/4.
2018. https://doi.org/10.1007/s11042-018-5675-4
[5] Hua Z., Zhang K., Li Y., and Zhou Y., “Visually
Figure 12. Cropping attack analysis of Clock image using different
degrees of occlusion.
Secure Image Encryption Using Adaptive-
Image Encryption Using Modified Perturbed Logistic Map 877

Thresholding Sparsification and Parallel https://doi.org/10.1007/s11042-019-07808-6

Compressive Sensing,” Signal Processing, vol. [16] Shakiba A., “A Novel Randomized Bit-Level
183, pp. 107998, 2021. Two-Dimensional Hyperchaotic Image
https://doi.org/10.1016/j.sigpro.2021.107998 Encryption Algorithm,” Multimedia Tools and
[6] Jiang X., Jiang G., Wang Q., and Shu D., “Image Applications, vol. 79, no. 43-44, pp. 32575-32605,
Encryption Algorithm Based on 2D‐CLICM 2020. https://doi.org/10.1007/s11042-020-09434-z
Chaotic System,” IET Image Processing, vol. 17, [17] Sheela S., Sanjay A., Suresh K., Tandur D., and
no. 7, pp. 2127-2141, 2023. Shubha G., “Image Encryption Based on 5D
https://doi.org/10.1049/ipr2.12778 Hyperchaotic System Using Hybrid Random
[7] Khanzadi H., Eshghi M., and Borujeni S., “Image Matrix Transform,” Multidimensional Systems
Encryption Using Random Bit Sequence Based on and Signal Processing, vol. 33, no. 2, pp. 579-595,
Chaotic Maps,” Arabian Journal for Science and 2022. https://doi.org/10.1007/s11045-021-00814-8
Engineering, vol. 39, no. 2, pp. 1039-1047, 2014. [18] Wang Q., Zhang X., and Zhao X., “Image
https://doi.org/10.1007/s13369-013-0713-z Encryption Algorithm Based on Improved Zig
[8] Liu Y., Qin Z., Liao X., and Wu J., “A Chaotic Zag Transformation and Quaternary DNA
Image Encryption Scheme Based on Hénon- Coding,” Journal of Information Security and
Chebyshev Modulation Map and Genetic Applications, vol. 70, pp. 103340, 2022.
Operations,” International Journal of Bifurcation https://doi.org/10.1016/j.jisa.2022.103340
and Chaos, vol. 30, no. 6, pp. 2050090, 2020. [19] Wang X. and Si R., “A New Chaotic Image
https://doi.org/10.1142/S021812742050090X Encryption Scheme Based on Dynamic L-Shaped
[9] Man Z., Li J., Di X., Sheng Y., and Liu Z., Scrambling and Combined Map Diffusion,”
“Double Image Encryption Algorithm Based on Optik, vol. 245, pp. 167658, 2021.
Neural Network and Chaos,” Chaos, Solitons and https://doi.org/10.1016/j.ijleo.2021.167658
Fractals, vol. 152, pp. 111318, 2021. [20] Wang X. and Zhang M., “An Image Encryption
https://doi.org/10.1016/j.chaos.2021.111318 Algorithm Based on New Chaos and Diffusion
[10] Patel S., Bharath K., and Kumar R., “Symmetric Values of a Truth Table,” Information Sciences,
Keys Image Encryption and Decryption Using 3D vol. 579, pp. 128-149, 2021.
Chaotic maps with DNA Encoding Technique,” https://doi.org/10.1016/j.ins.2021.07.096
Multimedia Tools and Applications, vol. 79, no. [21] Wang X. and Zhao M., “An Image Encryption
43-44, pp. 31739-31757, 2020. Algorithm Based on Hyperchaotic System and
https://doi.org/10.1007/s11042-020-09551-9 DNA Coding,” Optics and Laser Technology, vol.
[11] Patro K., Soni A., Netam P., and Acharya B., 143, pp. 107316, 2021.
“Multiple Grayscale Image Encryption Using https://doi.org/10.1016/j.optlastec.2021.107316
Cross-Coupled Chaotic Maps,” Journal of [22] Wang X., Guan N., and Yang J., “Image
Information Security and Applications, vol. 52, Encryption Algorithm with Random Scrambling
pp. 102470, 2020. Based on One-Dimensional Logistic Self-
https://doi.org/10.1016/j.jisa.2020.102470 Embedding Chaotic Map,” Chaos, Solitons and
[12] Ponuma R. and Amutha R., “Image Encryption Fractals, vol. 150, pp. 111117, 2021.
Using Sparse Coding and Compressive Sensing,” https://doi.org/10.1016/j.chaos.2021.111117
Multidimensional Systems and Signal Processing, [23] Wang X., Liu L., and Zhang Y., “A Novel Chaotic
vol. 30, no. 4, pp. 1895-1909, 2019. Block Image Encryption Algorithm Based on
https://doi.org/10.1007/s11045-019-00634-x Dynamic Random Growth Technique,” Optics
[13] Ponuma R. and Amutha R., “Compressive and Lasers in Engineering, vol. 66, pp. 10-18,
Sensing Based Image Compression-Encryption 2015. https://doi.org/10.1016/j.optlaseng.2014.08.005
Using Novel 1D-Chaotic Map,” Multimedia Tools [24] Wang X., Zhang Y., and Bao X., “A Novel
and Applications, vol. 77, no. 15, pp. 19209- Chaotic Image Encryption Scheme Using DNA
19234, 2018. https://doi.org/10.1007/s11042-017- Sequence Operations,” Optics and Lasers in
5378-2 Engineering, vol. 73, pp. 53-61, 2015.
[14] Ponuma R. and Amutha R., “Encryption of Image https://doi.org/10.1016/j.optlaseng.2015.03.022
Data Using Compressive Sensing and Chaotic [25] Wei D., Jiang M., and Deng Y., “A Secure Image
System,” Multimedia Tools and Applications, vol. Encryption Algorithm Based on Hyper-Chaotic
78, no. 9, pp. 11857-11881, 2019. and Bit-Level Permutation,” Expert Systems with
https://doi.org/10.1007/s11042-018-6745-3 Applications, vol. 213, pp. 119074, 2023.
[15] Ponuma R., Amutha R., Aparna S., and Gopal G., https://doi.org/10.1016/j.eswa.2022.119074
“Visually Meaningful Image Encryption Using [26] Wu Y., Zhou Y., Saveriades G., Agaian S.,
Data Hiding and Chaotic Compressive Sensing,” Noonan J., and Natarajan P., “Local Shannon
Multimedia Tools and Applications, vol. 78, no. Entropy Measure with Statistical Tests for Image
18, pp. 25707-25729, 2019. Randomness,” Information Sciences, vol. 222, pp.
878 The International Arab Journal of Information Technology, Vol. 21, No. 5, September 2024

323-342, 2013. Subhashini Kumaran obtained her

https://doi.org/10.1016/j.ins.2012.07.049 UG degree in ECE from Madras
[27] Xian Y. and Wang X., “Fractal Sorting Matrix and University in the year 2001 and PG
its Application on Chaotic Image Encryption,” degree in Medical Electronics from
Information Sciences, vol. 547, pp. 1154-1169, College of Engineering, Guindy,
2021. https://doi.org/10.1016/j.ins.2020.09.055 Chennai in the year 2007. She is
[28] Xu C., Sun J., and Wang C., “An Image currently working in Sri Sai Ram
Encryption Algorithm Based on Random Walk Engineering College, Chennai, India. She is a senior
and Hyperchaotic Systems,” International member of IEEE, fellow member of IETE and life
Journal of Bifurcation and Chaos, vol. 30, no. 4, member of ISTE. Her current research interest includes
pp. 2050060, 2020. Image Processing.
https://doi.org/10.1142/S0218127420500601
[29] Xu D., Li G., Xu W., and Wei C., “Design of Amutha Ramachandran is a
Artificial Intelligence Image Encryption Professor in the Department of ECE,
Algorithm Based on Hyperchaos,” Ain Shams Sri Sivasubramaniya Nadar College
Engineering Journal, vol. 14, no. 3, pp. 101891, of Engineering, Chennai, India.
2023. https://doi.org/10.1016/j.asej.2022.101891 Received her UG degree in ECE from
[30] Yang F., Mou J., Cao Y., and Chu R., “An Image Thiagarajar college of Engineering,
Encryption Algorithm Based on BP Neural India, in 1987. Received her PG
Network and Hyperchaotic System,” China degree from PSG college of Technology, Coimbatore,
Communications, vol. 17, no. 5, pp. 21-28, 2020. India and she was awarded PhD degree in the year 2006
https://doi.org/10.23919/JCC.2020.05.003 from Anna University. Her total teaching experience is
[31] Ye G., Pan C., Dong Y., Shi Y., and Huang X., 37 years which includes 20 years of research
“Image Encryption and Hiding Algorithm Based experience. Her research interest includes Wireless
on Compressive Sensing and Random Numbers Communication Network and Image Processing.
Insertion,” Signal Processing, vol. 172, pp.
107563, 2020.
https://doi.org/10.1016/j.sigpro.2020.107563
[32] Ye G., Wu H., Jiao K., and Mei D., “Asymmetric
Image Encryption Scheme Based on the Quantum
Logistic Map and Cyclic Modulo Diffusion,”
Mathematical Biosciences and Engineering, vol.
18, no. 5, pp. 5427-5448, 2021.
DOI:10.3934/mbe.2021275
[33] Zhang Q., Han J., and Ye Y., “Multi‐Image
Encryption Algorithm Based on Image Hash, Bit‐
Plane Decomposition and Dynamic DNA
Coding,” IET Image Processing, vol. 15, no. 4, pp.
885-896, 2021.
https://doi.org/10.1049/ipr2.12069
[34] Zhang Q., Liu L., and Wei X., “Improved
Algorithm for Image Encryption Based on DNA
Encoding and Multi-Chaotic Maps,” AEU-
International Journal of Electronics and
Communications, vol. 68, no. 3, pp. 186-192,
2014. https://doi.org/10.1016/j.aeue.2013.08.007