Arnab - Image Encryption Based On 8D Hyperchao
Arnab - Image Encryption Based On 8D Hyperchao
Keywords: One of the key tenets of the enormous and still-expanding digital realm is the security of images. In recent
Hyperchaotic system years, the interest in image encryption has increased among researchers. This paper proposes an image
Fibonacci Q-Matrix encryption scheme using an 8D hyperchaotic system with Fibonacci Q-Matrix (FQ-matrix). The strength of
Encryption
the proposed image encryption algorithm is estimated by checking the statistical attacks (Histogram analysis,
Statistical attacks
correlation coefficient, and information entropy) and differential attacks (NPCR, UACI).
Differential attacks
∗ Corresponding author.
E-mail addresses: [email protected] (G. Biban), [email protected] (R. Chugh), [email protected] (A. Panwar).
https://doi.org/10.1016/j.chaos.2023.113396
Received 28 November 2022; Received in revised form 15 March 2023; Accepted 20 March 2023
Available online 27 March 2023
0960-0779/© 2023 Published by Elsevier Ltd.
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 1. MATLAB 2D and 3D phase plots of the 8D hyperchaotic system for 𝑋(0) = (1, 1, 1, 1, 0, 0, 0, 0) and (𝜏1 , 𝜏2 , 𝜏3 , 𝜏4 , 𝜏5 , 𝜏6 , 𝜏7 ) = (10, 76, 3, 0.2, 0.1, 0.1, 0.2).
In Section 2, Preliminaries contains the mathematical fundamentals system. Fig. 2 represents the system has six positive Lyapunov expo-
of the 8D hyperchaotic map and FQ-matrix. The algorithm for encrypt- nents which indicate that the system is hyperchaotic.
ing and decrypting an image are in Section 3. Experimental results show
the encrypted and decrypted images based on the proposed algorithms 2.2. Fibonacci Q-Matrix [26]
in Section 4. Statistical analysis like histogram, correlation coefficient,
and entropy analysis is described in Section 5. Section 6 contains the
The elements of the Fibonacci sequence 𝐹𝑛 are
differential attacks. At last, this paper is concluded in Section 7.
𝐹𝑛 = 𝐹𝑛−1 + 𝐹𝑛−2 ; 𝑛 > 1,
2. Preliminaries
where 𝐹1 = 𝐹2 = 1. The FQ-matrix is given by:
[ ] [ ]
2.1. 8D hyperchaotic system [25] 𝐹 𝐹1 1 1
𝑄= 2 = ,
𝐹1 𝐹0 1 0
Let 𝑋 = (𝑥1 , 𝑥2 , 𝑥3 , 𝑥4 , 𝑥5 , 𝑥6 , 𝑥7 , 𝑥8 ) denotes the state of the 8D
where 𝐹𝑛 is a Fibonacci number and |𝑄| = −1. Then
hyperchaotic system.
[ ]
⎧ 𝐹 𝐹𝑛
𝑄𝑛 = 𝑛+1 ,
⎪𝑥̇1 = 𝜏1 (𝑥2 − 𝑥1 ) + 𝑥4 ; 𝐹𝑛 𝐹𝑛−1
⎪𝑥̇ = 𝜏2 𝑥1 − 𝑥1 𝑥3 + 𝑥4 ;
⎪ 2
⎪𝑥̇3 |𝑄𝑛 | = 𝐹𝑛+1 𝐹𝑛−1 − 𝐹𝑛2 = (−1)𝑛 ,
= 𝑥1 𝑥2 − 𝑥3 − 𝑥4 + 𝑥7 ;
⎪
⎪𝑥̇4 = −𝜏3 (𝑥1 + 𝑥2 ) + 𝑥5 ; [ ]
⎨ (1) 𝐹𝑛−1 −𝐹𝑛
= −𝑥2 − 𝜏4 𝑥4 + 𝑥6 ; 𝑄−𝑛 = .
⎪𝑥̇5 −𝐹𝑛 𝐹𝑛+1
⎪𝑥̇ = −𝜏5 (𝑥1 + 𝑥5 ) + 𝜏4 𝑥7 ;
⎪ 6
⎪𝑥̇7 = −𝜏6 (𝑥1 + 𝑥6 − 𝑥8 ); 3. Algorithm
⎪
⎪𝑥̇8 = −𝜏7 𝑥7 ,
⎩ 3.1. Encryption process
where (𝜏1 , 𝜏2 , 𝜏3 , 𝜏4 , 𝜏5 , 𝜏6 , 𝜏7 ) are the seven constant parameters deter-
mining the chaotic, hyperchaotic, periodic behaviors, and bifurcations Image encryption algorithm are based on two key steps. Confusion
of the system. The system exhibits hyperchaotic behavior when 𝑋(0) = is the first step, in which the 8D hyperchaotic system changes the pixel
(1, 1, 1, 1, 0, 0, 0, 0) and (𝜏1 , 𝜏2 , 𝜏3 , 𝜏4 , 𝜏5 , 𝜏6 , 𝜏7 ) = (10, 76, 3, 0.2, 0.1, 0.1, 0.2). arrangements. The second step is diffusion, in which the FQ-matrix
The 8D hyperchaotic system has eight Lyapunov exponents (𝐿𝐸 ′ ) are : changes the value of pixels.
𝐿𝐸1′ = 1.4565, 𝐿𝐸2′ = 0.1176, 𝐿𝐸3′ = 0.0623, Step 1: Given plain image 𝐷 with size 𝑀 × 𝑁 and convert image 𝐷 to
𝐿𝐸4′ = 0.0433, 𝐿𝐸5′ = 0.0260, 𝐿𝐸6′ = 0.0132, vector 𝑆.
𝐿𝐸7′ = 0 and 𝐿𝐸8′ = −12.6987. Step 2: The confusion stage depends on the 8D hyperchaotic map. We
determine the initial key of the hyperchaotic system using
Fig. 1 shows the 2D and 3D phase plots of the 8D hyperchaotic system
∑𝑀𝑁
at 𝑋(0) = (1, 1, 1, 1, 0, 0, 0, 0) and (𝜏1 , 𝜏2 , 𝜏3 , 𝜏4 , 𝜏5 , 𝜏6 , 𝜏7 ) = (10, 76, 3, 0.2, 𝑆(𝑖) + (𝑀 × 𝑁)
0.1, 0.1, 0.2) for better understanding the nature of the 8D hyperchaotic 𝑥(1) = 𝑖=1 ,
223 + (𝑀 × 𝑁)
2
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 2. LE’s of the 8D hyperchaotic system for 𝑋(0) = (1, 1, 1, 1, 0, 0, 0, 0) and (𝜏1 , 𝜏2 , 𝜏3 , 𝜏4 , 𝜏5 , 𝜏6 , 𝜏7 ) = (10, 76, 3, 0.2, 0.1, 0.1, 0.2).
3
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 4. (a) Plain image (b) Encrypted image (c) Decrypted image.
Where 𝑌 is the solution array of the 8D hyperchaotic system using 3.2. Decryption process
MATLAB function ode45.
In this section, the reverse procedure of encryption is carried out.
Step 5: The diffusion step depends on FQ-matrix. In this disordered
We use the 8D hyperchaotic map to implement the decryption process.
image, 𝑃 is convert into 𝑀 × 𝑁 matrix and is separated into subblocks
of size 2 × 2. Step 1: Read cipher image 𝐶 and separate it into subblocks of size 2×2.
We use the following diffusion equation with 𝑄−10 on image blocks to
Step 6: Using FQ-matrix multiply each subblock with 𝑄10 matrix then
obtain scrambled image 𝐾 ′ .
get matrix 𝐶 and plain image 𝐷 = 𝐶. Get the output in cipher image or [ ]
′
𝐾𝑚,𝑛 ′
𝐾𝑚,𝑛+1
encrypted image in the form 𝐶. The proposed methodology to encrypt
′ ′ =
an image is given in Fig. 3. 𝐾𝑚+1,𝑛 𝐾𝑚+1,𝑛+1
4
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 5. Histogram analysis corresponding to (a) Plain image (b) Encrypted image (c) Decrypted image according to Fig. 4.
[ ][ ]
𝐶𝑚,𝑛 𝐶𝑚,𝑛+1 34 −55
𝑚𝑜𝑑 256,
𝐶𝑚+1,𝑛 𝐶𝑚+1,𝑛+1 −55 89
Step 4: Convert 𝑄 into matrix 𝐹 and 𝐶 = 𝐷 and get the decrypted
where 𝑚 = 1 ∶ 3 ∶ 5 ∶ ... ∶ 𝑀, and 𝑛 = 1 ∶ 3 ∶ 5 ∶ ... ∶ 𝑁. image 𝐷.
Step 2: Convert image 𝐾 ′ into vector 𝑊 .
4. Experimental results
Step 3: The vector 𝐿 generated in the encryption procedure is used to
restore each pixel back to its original place by the following equation:
Different gray images, such as ‘‘Lena’’, ‘‘Baboon’’, ‘‘Peppers’’, and
𝑄(𝐿𝑚 ) = 𝑊𝑚 ; 𝑚 = 1 ∶ 𝑀𝑁. ‘‘Airplane’’ with size 256 × 256 were used for research. Fig. 4 shows
5
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 6. Correlation analysis (a) Lena plain image (b) Lena encrypted image.
the results of encryption and decryption images. We have used Mat- 5. Statistical analysis
lab 9.10.0 (R2021a) to run the proposed encryption and decryption
algorithm in a computer including an Intel (R) Core i5 CPU 1.60 GHz, Numerous tests have been conducted to demonstrate the algo-
256 GB SSD, and 1 TB hard-disk capacity, and the operating system is rithm’s resistance to statistical attacks, including histogram, correlation
Microsoft Windows 11 version 21 H2. coefficient, and entropy analysis.
6
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 7. Correlation analysis (a) Baboon plain image (b) Baboon encrypted image.
1 ∑
𝑇 5.3. Entropy analysis
𝐷(𝑠) = (𝑠 − 𝐸(𝑠))2 ,
𝑇 𝑖=1 𝑖
Entropy is an essential tool to analyze image unpredictability. The
where 𝐸(𝑠) and 𝐷(𝑠) represent the expectation and variance of 𝑠, distribution of gray values in the image is quantified by entropy.
respectively and the total number of adjacent pixels are represented Entropy is defined as:
by 𝑇 and 𝑠𝑖 , 𝑡𝑖 are values of adjacent pixels in the image. When the
𝑤 −1
2∑
absolute value of the correlation coefficient is smaller, it indicates that 1
𝐻(𝑔) = 𝑃 (𝑔𝑖 ) log2 ,
the correlation between pixels is less. 𝑃 (𝑔𝑖 )
𝑖=1
Figs. 6–9 visualize the change of correlation before and after en-
cryption of the ‘‘Lena’’, ‘‘Baboon’’, ‘‘Peppers’’ and ‘‘Airplane’’ images, where the occurrence probability of 𝑔𝑖 is 𝑃 (𝑔𝑖 ) and 2𝑤 shows the total
respectively. The figures manifest that there is a linear correlation number of 𝑔𝑖 , where 𝑤 represents the total number of pixels in the
between adjacent pixels of the plain image and uniform distribution image. An optimum value of entropy for the gray level images is eight.
of the adjacent pixels for the encrypted image that represents almost Table 2 provides the entropy of the different plain images and the
no correlation between adjacent pixels. The correlation coefficients for respective cipher images encrypted by different schemes. The result
the numerous plain images and their cipher images, which had been en- demonstrates that most of the entropy values of our proposed algorithm
crypted using various schemes, are shown in Table 1 for the horizontal, are significantly closer to the optimum entropy value.
vertical, and diagonal directions. In the successfully encrypted images,
the correlation between adjacent pixels should approach 0. Results 6. Differential attacks
indicate that the correlation between adjacent pixels in the encrypted
image can be removed by the proposed algorithm. In conclusion, The differential attacks include the Number of Pixel Change Rate
the correlation coefficient values are statistically insignificant, i.e. the (NPCR) and the Unified Average Changing Intensity (UACI).
7
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Table 1
Comparative correlation analysis of pixels in three directions: Horizontal (H), Vertical (V), and Diagonal (D) for images with our algorithm and other encryption algorithms.
Average
Table 2
Comparative analysis of entropy values for images with proposed algorithm and other encryption algorithms.
Average
Table 3
Comparative analysis of NPCR, UACI values with proposed algorithm and other encryption algorithms.
Average
8
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 8. Correlation analysis (a) Peppers plain image (b) Peppers encrypted image.
NPCR is used to measure the change in the number of pixels in gray 7. Conclusion
level in two images. The encrypted images 𝐸 and 𝐸 ′ are encrypted from
an original image with only one bit dissimilar. Let 𝐸(𝑖, 𝑗) and 𝐸 ′ (𝑖, 𝑗) This paper implemented an image encryption method using an
be pixels at the location 𝑖th row and 𝑗th column of two images 𝐸 and FQ-matrix and 8D hyperchaotic system. Firstly, the confusion step is
𝐸 ′ , respectively. The total number of pixels represented by 𝑃 . NPCR is performed to generate random sequences using an 8D hyperchaotic
defined as: system then three of these sequences are chosen to shift the pixel
∑ placement of the gray images. In diffusion operation, each sub-block
𝑖,𝑗 𝐶(𝑖, 𝑗)
𝑁𝑃 𝐶𝑅 = × 100, of the shuffled image has its pixel value which we have changed using
𝑃
the FQ-matrix.
where, The advantage of the proposed image encryption algorithm is that
{
0 if 𝐸(𝑖, 𝑗) = 𝐸 ′ (𝑖, 𝑗), it is based on 8D hyperchaotic system which contains twenty-three
𝐶(𝑖, 𝑗) = terms on the right-hand side and seven control parameters giving large
1 if 𝐸(𝑖, 𝑗) ≠ 𝐸 ′ (𝑖, 𝑗).
key space and providing a very secure image encryption algorithm and
UACI is used to obtain the average change in intensities of two images. generated encryption algorithm produces low time complexity. For the
UACI is defined as: encrypted images, the histogram distribution is uniform, and the corre-
1 ∑ |𝐸(𝑖, 𝑗) − 𝐸 ′ (𝑖, 𝑗)| lation coefficient between adjacent pixels has significantly decreased.
𝑈 𝐴𝐶𝐼 = × 100.
𝑃 𝑖,𝑗 255 Entropy measurement demonstrates that the encrypted image has a
high level of randomness and is extremely close to the optimum entropy
The encryption algorithm is superior when the NPCR and UACI value. The NPCR and UACI values of the proposed encryption algorithm
value is closer to the ideal NPCR and UACI value i.e. 99.6094 and are much closer to the ideal values. We may infer from the analysis
33.4635, respectively. Table 3 represents the NPCR and UACI com- findings that the proposed algorithm provided a good level of security
puted values of the proposed algorithm and the other existing image that facilitates it to protect gray images in the public domain from
encryption algorithms. hackers.
9
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
Fig. 9. Correlation analysis (a) Airplane plain image (b) Airplane encrypted image.
10
G. Biban et al. Chaos, Solitons and Fractals: the interdisciplinary journal of Nonlinear Science, and Nonequilibrium and Complex Phenomena 170 (2023) 113396
[14] Sheela SJ, Sanjay A, Suresh KV, Tandur D, Shubha G. Image encryption based [21] Kumar A, Alzabut J, Kumari S, Rani M, Chugh R. Dynamical properties of a
on 5D hyperchaotic system using hybrid random matrix transform. Multidimens novel one dimensional chaotic map. Math Biosci Eng 2022;19(3):2489–505.
Syst Signal Process 2022;33(2):579–95. [22] Kumari S, Chugh R. A new experiment with the convergence and stability of
[15] Gong L, Deng C, Pan S, Zhou N. Image compression-encryption algorithms logistic map via SP orbit. Int. J. Appl. Eng. Res. 2019;14:797–801.
by combining hyper-chaotic system with discrete fractional random transform. [23] Kumari S, Chugh R, Miculescu R. On the complex and chaotic dynamics of
Optics Laser Technol 2018;103:48–58. standard logistic Sine square map. An. ŞtiinŢ. Univ. ‘‘Ovidius’’ ConstanŢa Ser.
[16] Gonzalez RC. Digital Image Processing. third ed.. Pearson education India; 2009. Mat. 2021;29(3):201–27.
[17] Gonzalez RC. Digital Image Processing using MATLAB. Pearson education India; [24] Rani M, Kumar A, Kumari S, Chugh R. Superior tent map. Nonlinear Studies
2004. 2021;28(3).
[18] Batra C, Gupta N, Chugh R, Kumar R. Generalized viscosity extragradient [25] Benkouider K, Bouden T, Yalcin ME, Vaidyanathan S. A new family of 5d, 6d,
algorithm for pseudomonotone equilibrium and fixed point problems for finite 7D and 8D hyperchaotic systems from the 4D hyperchaotic vaidyanathan system,
family of demicontractive operators. J Appl Math Comput 2022;1–28. the dynamic analysis of the 8D hyperchaotic system with six positive Lyapunov
[19] Geeta, Panwar A. Data analysis of COVID-19 pandemic: a mathematical exponents and an application to secure communication design. Int J Model Ident
approach. J. Math. Comput. Sci. 2021;11(6):7568–623. Control 2020;35(3):241–57.
[20] Gupta N, Postolache M, Nandal A, Chugh R. A cyclic iterative algorithm for [26] Hoggatt VE. Fibonacci and Lucas Numbers. Houghton Mifflin; 1969.
multiple-sets split common fixed point problem of demicontractive mappings
without prior knowledge of operator norm. Mathematics 2021;9(4):372.
11