IETE Technical Review

ISSN: 0256-4602 (Print) 0974-5971 (Online) Journal homepage: www.tandfonline.com/journals/titr20

Image Encryption using DNA Coding and

Hyperchaotic System

M. Kar, A. Kumar, D. Nandi & M. K. Mandal

To cite this article: M. Kar, A. Kumar, D. Nandi & M. K. Mandal (2020) Image Encryption
using DNA Coding and Hyperchaotic System, IETE Technical Review, 37:1, 12-23, DOI:
10.1080/02564602.2018.1544855

To link to this article: https://doi.org/10.1080/02564602.2018.1544855

Published online: 20 Nov 2018.

Submit your article to this journal

Article views: 251

View related articles

View Crossmark data

Citing articles: 11 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=titr20
IETE TECHNICAL REVIEW
2020, VOL. 37, NO. 1, 12–23
https://doi.org/10.1080/02564602.2018.1544855

Image Encryption using DNA Coding and Hyperchaotic System

M. Kar1 , A. Kumar1 , D. Nandi2 and M. K. Mandal1
1 Department of Physics, National Institute of Technology, Durgapur, India; 2 Department of Information Technology, National Institute of
Technology, Durgapur, India

ABSTRACT KEYWORDS
The paper proposes a grey scale image encryption scheme using six-dimensional Lorenz chaotic Lorenz system; Chaos;
system. The key generating system uses a 32 character (256 bits) key as input. The system uses Encryption; Decryption; DNA
simple modular arithmetic and bitwise XOR operations followed by coupling a series of Lorenz coding; Information entropy;
six-dimensional systems to generate the key-stream. This generated key stream is used in the NPCR; UACI
encryption and decryption algorithm of the plain image. The basic permutation and diffusion
architecture, using the fundamental properties of modular arithmetic, bitwise XOR operations and
DNA encoding/decoding technique, are executed to confuse the relationship between plain image
and the ciphered image. The experimental results and security tests confirm that the proposed
image encryption algorithm possess high security and can be a good candidate for practical image
encryption.

1. INTRODUCTION variables and fixed parameters may result in a sequence

which is deterministic but unpredictable when iterated
Now-a-days data transmission is prominent in any sort
for a number of times. The generated sequence exhibits
of communication and the importance of a particular
an inherent behavior of randomness and this property
type of data can vary from lower to higher level. In the
is widely exploited in designing the chaos based cryp-
present age of technology, cybercrime is quite common
tosystems. In the literature of chaos-based cryptography,
and the security of data may easily be compromised. In
generally, four types of chaotic systems (discrete time,
order to counter the unauthorised access and data theft,
continuous time, time delay and hyperchaotic) used as
a suitable data protection technique should be developed.
entropy sources in an encryption algorithm. The role of
Data encryption is one of the important data protection
the chaotic system in encryption algorithm is to gener-
techniques which widely used in information security.
ate the chaotic key sequence for pixel permutation and
Image data encryption has gained popularity since any
confusion. A few examples of chaos based cryptosystems
type of data can be represented in the form of an image.
along with their merits and demerits are discussed in the
Conventional cryptographic algorithms like AES, DES,
next section.
may be quite successful in encrypting raw data of small
size but cannot provide good results in encryption of the
digital images. These algorithms are unable to exploit
2. RELATED LITERATURE REVIEW
the correlation among the adjacent pixels, which is an
inherent property of an image. Rigorous research is going Hong et al. have proposed a suitable chaos based
on the techniques of encrypting an image. Any stan- using two round encryption with two different chaotic
dard encryption algorithm performs basic three opera- sequences generated from the logistic map and hyper-
tions: pixel permutation, diffusion and confusion. How- chaotic system, respectively, in ref. [1]. Min et al. com-
ever, as the time passed by, these encryption algorithms bined a piecewise linear chaotic map with a logistic
started to fall prey to certain stringent security test. Desir- map in reference [2] to produce a chaotic pseudo ran-
able randomness in the generated sequence could not dom number sequence and applied encrypting image.
be obtained from these algorithms. Hence non-linear Lin and Wang have proposed an chaotic image encryp-
science started to gain popularity. Chaos theory intro- tion method by using the basic process of image scram-
duced by Edward Lorenz became the building block of bling, pixel replacement and piecewise linear memristor
chaos based image encryption techniques. A certain set in Chua’s circuit in reference [3]. An image encryption
of deterministic equations composed of a number of state algorithm using chaotic Baker map and Lorenz system is

© 2020 IETE
 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM 13

introduced in reference [4]. It proposes an image encryp- information entropy of the cryptosystem. A spatiotem-
tion algorithm in which the key stream is dependent poral chaos based on coupled map lattice is used in ref.
on the input plain image. Further, the state variables of [12] to generate chaotic sequence and pixel confusion by
the Lorenz system are selected according to the pixel XOR operation. Then DNA encoding followed by posi-
values of the plain image. Reference [5] proposes an tion permutation is done. Then a specific rule based DNA
image encryption technique with orthogonal polyno- decoding is performed to generate the ciphered image.
mial transformation in the frequency domain and bit Other noteworthy algorithms are discussed in references
shuffling using tent map in the spatial domain. In ref- [13–17]. Some of the algorithms stated above do not give
erence [6] an image encryption algorithm employing good results in various security analysis or not technically
Latin rectangle scrambling using 2D chaotic Henon map sound in the view point of computational complexity and
and bi-directional pixel diffusion using piecewise linear execution time.
chaotic map is proposed. The algorithm claims to reach
high security in only two rounds. A novel double image Therefore, the aim of this paper is to propose a new
encryption technique using Blowfish algorithm and cross encryption system that can reduce the complexity of the
chaos map is presented in reference [7]. A dynamical encryption algorithm by considering only one hyper-
image encryption algorithm based on parameter vary- chaotic system for pixel confusion, permutation, DNA
ing logistic chaotic map is presented in reference [8]. encoding, complementing and decoding to achieve the
Reference [9] has presented a new block based image ciphered image with higher security. In this algorithm, a
encryption. It proposes to use cat map as well as logistic six-dimensional chaotic Lorenz system is used to design
map in encryption process. The input image dependent a cryptosystem having reduced complexity and execu-
key stream generated from these maps are used in the tion time compared to the reported work in the literature
diffusion process. The algorithm claims secured enough recently. A key matrix is generated by taking a 32 char-
in practical transmission and reception. A colour image acter secret key as input. The generated key matrix is
encryption algorithm using 4D Lorenz system is reported reshaped into a matrix with the same dimension of the
from our group in an international conference [10]. In input image. The pixel positions of the plain image are
this paper, the colour image is spleet into three R G B randomized using chaotic key matrix. After XOR opera-
channel and they are encrypted by the mechanism of tion between plain image and chaotic key, DNA encod-
pixel diffusion and confusion with the help of chaotic key ing, complement shifting and DNA decoding are applied
generated from 4D Lorenz system. Next pixel permuta- sequentially to execute the encryption process. The rest
tion is done among three colour components with the of the paper is organized as follows: Section 3 outlines
help of chaotic quadratic and cubic maps. In the above the 6D Lorenz chaotic system and its relevant character-
mentioned references, the encryption algorithms have istics. The basic rules for DNA encoding and decoding
involved two round operations and have used double or techniques are illustrated in Section 4 along with suit-
multiple chaotic systems for chaotic key generation. Since able examples. Section 5 presents the grey scale image
a two round encryption takes longer execution time in encryption and decryption algorithms. Section 6 presents
comparison to a single round encryption and the inclu- the experimental results and the security analysis of the
sion of double or multiple chaotic systems increases the proposed cryptosystem and the article is concluded in
system complexity, we need to develop a fast and highly section 7.
secured chaotic encryption system with relatively less
system complexity.
3. LORENZ CHAOTIC SYSTEM
Recently, chaos based DNA coding and DNA addition- The proposed image cryptosystem uses a hyperchaotic
subtraction arithmetic are being used in several image six-dimensional Lorenz system [18] as stated below:
encryption algorithms [11–15] to resist the known- ⎫
ẋ = a(y − x) + u ⎪
⎪
ẏ = cx − xz − y + v ⎪
plaintext, brute force, differential and chosen-plaintext ⎪
⎪
⎪
attacks. In reference [11] an image encryption algorithm ⎬
ż = −bz + xy
utilizing chaotic systems, DNA encoding is presented. In (1)
u̇ = du − xz ⎪
⎪
this algorithm, the logistic system is iterated to produce ⎪
⎪
v̇ = −ky ⎪
⎪
the DNA matrix and set of encoding/decoding rules is ⎭
ẇ = hw + ly
framed. DNA addition operation is performed using the
generated DNA matrix and the DNA encoded image. The where x, y, z, u, v, w are the state variables and a, b, c, d, k,
DNA encoded image diffusion is done through the itera- h and l are the parametes. In our algorithm we have taken
tions of the logistic system to increase the sensitivity and a = 10, b = 8/3, c = 28, d = 2, k = 8.4, h = 1 and l = 1
14 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM

Figure 1: The projections of attractors of 6D Lorentz system. (a) In the x–y space, (b) In the x–z space, (c) In the z–y space.

as the parameter values. The projection of attractors is Table 1: DNA encoding and decoding rules.
shown in Figure 1. The 6D chaotic Lorenz system has Rule-1 Rule-2 Rule-3 Rule-4 Rule-5 Rule-6 Rule-7 Rule-8
been chosen to design the cryptosystem since it exhibits 00-A 00-A 00-T 00-T 00-C 00-C 00-G 00-G
the property of randomness with higher degree of unpre- 01-C 01-G 01-C 01-G 01-A 01-T 01-A 01-T
10-G 10-C 10-G 10-C 10-T 10-A 10-T 10-A
dictability. Its higher chaotic nature has made it better 11-T 11-T 11-A 11-A 11-G 11-G 11-C 11-C
suitable candidate for secure communication applica-
tions. The characteristic properties of this hyperchaotic
system are: (a) it is a six-dimensional autonomous sys- as listed in Table 1. For example, if the decimal value of
tem, (b) It has a simple algebraic structure with complex a pixel is 53, its 8-bit binary representation is 00110101,
dynamical behaviour, (c) The six Lyapunov exponents then it is encoded as a DNA sequence CGAA according to
of the system are, λ1 = −2.667, λ2 = −22.6858, λ3 = rule 5 in Table 1. Inversely, the pixel value can be decoded
0.3260, λ4 = 1, λ5 = 2 and λ6 = 11.3599. Since it has by DNA sequence, for example, DNA sequence TCTC is
four positive Lyapunov exponents and the sum of all the decoded as 10111011 by rule 7 and the corresponding
Lyapunov exponents are less than zero it can produce sta- decimal value is 187.
ble but unpredictable orbits which implies the generation
of chaos. This paper has developed a dynamic DNA coding pro-
cess in which the rules of Table 1 is selected according
to a chaotic matrix, say A. The elements of A are inte-
4. THE DNA ENCODING AND DECODING gers in the range 1–8. The pre-processed image may be
PROCESS represented by another matrix, say B where the elements
The Deoxyribonucleic acid (DNA) is a molecule that of B are integers in the range 1–255 for 8-bits binary rep-
contains the genetic information of all the living organ- resentation of each pixel. The dimension of A and B are
isms. The development and function of the organisms same (say 3 × 3, for an example). Thus, the dimension
depend on the structure of its DNA molecule. A DNA of the DNA encoded matrix (say C) will be 3 × 12, as
structure consists of four nucleic acid bases adenine (A), illustrated below by taking arbitrary values of A and B.
cytosine (C), guanine (G) and thymine (T). According ⎛ ⎞ ⎛ ⎞
to Watson-Crick base pairing rules [11,19], A and T are 3 7 5 78 157 246
complementary and G and C are complementary. In the A = ⎝4 1 8 ⎠ B = ⎝124 42 207 ⎠
binary representation 0 and 1 are complementary, simi- 5 6 2 19 127 213
larly, we can say 00 and 11 are complementary and also 01
⎛ ⎞
and 10 are complementary. By considering the comple- C T A G T A C A G G T C
mentary rule, we may use A, C, G and T to denote 00, 01, C = ⎝G A A T A G G G C G C C ⎠
10 and 11, respectively. This is an example of one possi- C A C G T G G G T G G G
ble representation and other representations are given in
Table 1. An image pixel value (8-bits binary number) may
5. THE PROPOSED CRYPTOSYSTEM
be represented by four binary numbers each consisting of
two bits. These numbers are mapped to their correspond- A grey scale image P(r × c) is taken as input where r and c
ing DNA sequence. Therefore, each pixel of the image is are the number of rows and columns, respectively. Con-
represented by four acid bases. The 8-bit pixel value can ventionally, P(i, j) represents the pixel value at position
be encoded in 24 possible ways using 4 bases but Watson- (i, j). The proposed cryptosystem is divided into the fol-
Crick base pairing rules support only 8 representations lowing major steps: (i) generation of initial conditions
 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM 15

of Lorenz system from 32 character secret input key, (ii) Here, t(x) denotes the element value of matrix t at posi-
Generation of chaotic key matrix R and reshaping the tion x, abs(x) returns the absolute value of the number x,
chaotic key matrix R to S (see step-2), (iii) DNA encoding k(t(x)) denotes the element value of matrix k at position
and decoding. The stepwise image encryption procedure t(x). The operator ⊕ indicates bitwise XOR operation.
is described below: After the completion of the loop we redefine p and q as
p = m0 + p0 and q = n0 + q0. We use the values of p
Step-1: A user defined function takes 32 characters secret and q to construct a one-dimensional matrix, C(1 × 6) =
p q
key and the dimension of the plain image as input and linspace 1000 , 1000 , 6 . Here, linspace(a, b, n) returns a
returns the chaotic key matrix and variables p and q. one-dimensional matrix of n linearly spaced elements
The substeps described below explain the working of the ranging from atob. Initially one dimensional matrix,
function. Initially, we set six variables as m0 = 127, n0 = S0(1 × 6) = [xyzuvw] = IC, defines the six initial con-
127, p0 = 127, q0 = 127, p = 0 and q = 0. the secret key ditions for the Lorenz 6D system.
is converted into a double precision matrix k(1 × 32)
containing the ASCII value of each character. A loop is Step-2: Solve equation (1) by using the initial condition
defined to run for eight times to generate the new val- S0 and differential equation solver ode45. The iterative
ues of the variables m0, n0, p0 and q0. Within the loop process has been executed (n + 100) times and the results
the following operations are done sequentially. A one- are stored in matrix R((n + 100) × 6); where n = r × c.
dimensional matrix t is constructed in which the element To obtain the steady state chaotic solution, we discard
value ranges from 1 to 32. the sequence of first 100 iterations from the matrix R.
The rest part of the R matrix is renamed as S(n × 6).
Loop j To restrict the elements of S in the range 0–255 we have
applied the following operation (2). The matrix S(n × 6)
Z = 9-j; is stated below in equation (3).

t = [(4 ∗ j − 3)(4 ∗ z)(4 ∗ j − 2)(4 ∗ z − 1)]; S = mod((round(abs(S ∗ 105 ))), 256). (2)

m = 10 + mod((k(t(1)) + k(t(2))), abs(k(t(3))

− k(t(4))));
⎛ ⎞
x1 y1 z1 u1 v1 w1
n = 20 + mod((k(t(3)) + k(t(4))), abs(k(t(1)) ⎜ x2
⎜ y2 z2 u2 v2 w2 ⎟⎟
− k(t(2)))); ⎜ x3
⎜ y3 z3 u3 v3 w3 ⎟⎟
⎜ ··· ··· ··· ··· ··· ··· ⎟
p = 10 + mod((m + n), abs(m − n)); ⎜
S(n × 6) = ⎜ ⎟
⎜ ··· ··· ··· ··· ··· ··· ⎟
⎟
⎜xn−2 yn−2 zn−2 un−2 vn−2 wn−2 ⎟
q = 20 + mod((m + n), abs(n − m)); ⎜ ⎟
⎝xn−1 yn−1 zn−1 un−1 vn−1 wn−1 ⎠
xn yn zn un vn wn
c = mod((m ⊕ m0), 128);
(3)
d = mod((n ⊕ n0), 128);
The matrix S is used to produce six different chaotic key
e = mod((p ⊕ p0), 128); matrices Ki of dimension r × c by taking the consecutive
elements of S rowwise. Here i = 1, 2, 3, . . . 6.
f = mod(q ⊕ q0), 128);
Step-3: We pre-process the chaotic keys K1 and K2
m0 = m0 + c; according to the equation (4) to get the new random
sequences K12 .
n0 = n0 + d;
K12 = (K1 ⊕ K2 ). (4)
p0 = p0 + e;
The next two chaotic sequences K3 and K4 are used to get
q0 = q0 + f ; four random sequences Kr1 , Kr2 , Kr3 and Kr4 of dimen-
sion (r × c) by the following Equations (5)–(8) and they
End loop j. are combined to obtain a matrix Kr of dimension (r × 4c)
16 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM

by putting them side by side according to equation (9). element of KB is zero then the corresponding element of
DP will remain unchanged but if the element of KB is one
Kr1 = mod(K3 , 64) (5) then the corresponding element of DP will be replaced by
its complement. Finally, the elements of DP are decoded
Kr2 = mod(K4 , 64) (6) according to chaotic sequence K6 to get back the decimal
values of the pixels of the encrypted image EP.
Kr3 = mod((K3 + K4 ), 64) (7)
The decryption process is the reverse process of encryp-
Kr3 = mod(abs(K3 − K4 ), 64) (8) tion. During decryption step-1 to step-3 of encryp-
tion process should be executed sequentially and step-
4 should be executed in reverse order to get back the
Kr = [Kr1 Kr2 Kr3 Kr4 ]. (9)
original plain image.
By using matrix Kr and equation (10) stated below, we get
an binary array KB of dimension (r × 4c). 6. EXPERIMENTAL RESULTS AND SECURITY
 ANALYSIS
0, if Kr (i, j) ≥ 32
KB (i, j) = . (10)
0, if Kr (i, j) < 32 The 32 character secret key and different grey scale
plain images of size (r × c) are taken as inputs and
Here, i and j vary from 1 to r and 1 to 4c, respectively. their encryption is done. For example, the plain images
The elements of the remaining two chaotic sequences K5 like, Lena, Barbara, and Cameraman along with their
and K6 are converted into integers in the range 1–8 by the encrypted images, decrypted images with correct key
following Equations (11) and (12), respectively. ()+ < > (7790n,[]&ˆ%$% ∼ tqpd;3b.l@kx and slightly
different key (wrong key)()+ < > (7790n,[]&ˆ%$% ∼
K5 = mod(K5 , 8) + 1. (11)
tqpd;3b.l@ly are shown in Figure 2.

K6 = mod(K6 , 8) + 1. (12) A cryptosystem is considered to be strong and secured

Step-4: The first two chaotic keys K1 and K2 are used enough if it is able to pass well-defined security tests.
for shuffling the pixel positions of the input image. The There exit a number of performance and security mea-
pixels of the plain image P are randomized by swap- surement techniques such as key space and sensitivity,
ping the pixel positions between P(K1 (i, j), K2 (i, j)) and histogram, correlation coefficient, information entropy,
P(K2 (i, j), K1 (i, j)). Then the pixel shifted image matrix differential attack, noise attack, speed analysis etc. for
P and random sequences K12 are used to carry out the benchmarking a cryptosystem. The various tests have
following operations (13)–(16) and get the matrix XP of been done and the results obtained are presented in
dimension r × c. this section with a comparison to the others algorithms
r−1 c−1 reported in the literature recently.
i=2 j=2 P(i, j)
Pm = , (13)
(r − 2) × (c − 2) 6.1 Key Space Analysis

XP(1, 1) = P(1, 1) ⊕ K12 (1, 1) ⊕ Pm , (14) An inherent characteristic of a chaotic system is its high
sensitivity to initial conditions. A slight change in initial
condition may produce an entirely different sequence. In
XP(i, 1) = P(i, 1) ⊕ K12 (i, 1) ⊕ XP(i − 1, c), for i = 1 the present work, a novel use of 6D Lorenz system has
(15) been demonstrated. The 32 bit secret key goes through
a series of modular arithmetic and bitwise-XOR oper-
ations. Only one bit change in the input key produces
XP(i, j) = P(i, j) ⊕ K12 (i, j) ⊕ XP(i, j − 1), for j = 1 an entirely different chaotic sequence. The key space is
(16) 2256 which is sufficiently large to counter the brute force
attack.
In Equations (15) and (16), i and j vary from 1 to r and 1
to c, respectively. By applying DNA encoding rule given
6.2 Key Sensitivity Analysis
in Table 1, encode matrix XP(r, c) according to chaotic
sequence K5 to obtain DP of dimension (r × 4c) with To prove the validity of the proposed algorithm in terms
the DNA elements A, C, G and T. Next the elements of of key sensitivity, we encrypt a plain image of Lena using
DP are complemented according to the array KB . If the two slightly different keys as stated below:
 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM 17

Figure 3: The results of key sensitivity test. (a)

Plain image Lena, (b) Encrypted image with Key:
()+ < > (7790n,[]&ˆ%$% ∼ tqpd;3b.l@kx, (c) Encrypted
image with Key: ()+ < > (7790n,[]&ˆ%$% ∼ tqpd;3b.l@ky, (d)
Difference of (b) and (c).

Key-1: ()+ < > (7790n,[]&ˆ%$% ∼ tqpd;3b.l@kx.

Key-2: ()+ < > (7790n,[]&ˆ%$% ∼ tqpd;3b.l@ly.

Figure 3 describes the results of encrypting Lena

image with slightly different keys (only single bit
change between these two keys) and also the difference
between two encrypted images. The encrypted images
(Figure 3(b,c)) are remarkably different from one another
because their difference is also chaotic in nature. This
result proves that the encryption algorithm is highly sen-
sitive to encryption key.

6.3 Histogram Analysis

Histogram is an important tool for analysing statisti-
cal data. For an image, the histogram depicts how fre-
quently different pixel values occur. The maxima and
minima pixel value of an image falls in the range 0–255
for 8-bit representation. For a cryptographic algorithm
to be secure, the encrypted image histogram should
have an uniform distribution of pixel values so that
no one can extract any information of the plain image
from the histogram of the encrypted image. The his-
togram plot of the plain and the encrypted images are
Figure 2: Plain images (Lena, Cameraman and Barbara) and their shown in Figure 4. The histogram of three different plain
encryption and decryption results. Plain images (a), (e), (i); their images (Figure 4(a,c,e)) are remarkably different whereas
corresponding encrypted images (b), (f), (j); decrypted images the histogram of their corresponding encrypted images
with correct key (c), (g), (k); decrypted images with wrong key (d), (Figure 4(b,d,f)) are very much similar to one another.
(h), (l).
18 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM

Figure 4: Histogram plot of various plain images and their encrypted version. (a) Plain Lena, (b) Encrypted Lena, (c) Plain Cameraman,
(d) Encrypted Cameraman, (e) Plain Barbara, (f) Encrypted Barbara.

These results confirm that the encryption algorithm is Table 2: Correlation coefficients of various test images and
highly secure against histogram matching and analysis their encrypted counterparts.
based attack. Images Horizontal Vertical Diagonal
Plain Lena 0.9400 0.9693 0.9179
Encrypted Lena 0.0023 0.0013 –0.0067
Plain Cameraman 0.9335 0.9592 0.9087
6.4 Correlation Analysis Encrypted Cameraman 0.0069 0.0003 0.0052
Plain Barbara 0.9029 0.9400 0.8848
Another tool for measuring the security performance Encrypted Barbara 0.0031 0.0014 –0.0016
of a cryptosystem is correlation analysis. It is obvious
that the pixel correlation of plain image is high and this
analysis of correlation coefficients obtained from our pro-
value decreases sharply for encrypted images. To find out
posed method with other methods reported in the recent
pixel correlation, we take two adjacent pixels (in hori-
literature.
zontal, vertical and diagonal directions) from different
1 
places of the plain and encrypted images randomly. Then N
the pixel correlations are measured in these three direc- Ex = xi , (17)
N i=1
tions. In this experiment, we have taken 3500 pairs of
adjacent pixels randomly in horizontal, vertical and diag-
1 
N
onal directions from different positions of both the plain Dx = (xi − E(x))2 , (18)
and encrypted images. The computed values of the cor- N i=1
relation coefficient applying the formulas stated below
1 
N
are given in Table 2. The corresponding scattered plots cov(x, y) = (xi − E(x))(yi − E(y)). (19)
are shown in Figure 5. Table 3 presents a comparative N i=1
 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM 19

Figure 5: Correlation among two adjacent pixels. (a) Plain Lena in horizontal direction, (b) Plain Lena in vertical direction, (c) Plain Lena
in diagonal direction, (d) Encrypted Lena in horizontal direction, (e) Encrypted Lena in vertical direction, (f) Encrypted Lena in diagonal
direction.

Table 3: Comparative studies of correlation coefficients of 6.5 Information Entropy Analysis

encrypted Lena.
Encryption algorithm Horizontal Vertical Diagonal Information entropy is a method of measuring the degree
Ref. [3] –0.043970 –0.003190 –0.008500
of randomness of an information source. The pixel value
Ref. [4] –0.0013 –0.0274 –0.0199 of an image falls in the range 0–255. Hence it has 256
Ref. [8] 0.0021 0.0046 0.0033 states and the maximum entropy would ideally be 8
Ref. [10] 0.0214 0.0465 −0.0090
Ref. [12] 0.0011 0.0040 0.0080 which is possible if all the pixels have different grey lev-
Ref. [13] 0.0796 0.0166 0.0032 els. The information entropy, m is measured by using the
Ref. [15] –0.0029 0.0015 –0.0076
Ref. [18] –0.0230 0.0019 –0.0034 formula given below:
Proposed algorithm 0.0023 0.0013 –0.0067
N −1
2
1
E(m) = P(mi )log2 (21)
i=0
P(mi )
√ 
rxy = cov(x, y)/ Dx Dy , (20)
where P(mi ) is the probability of symbol mi and each
symbol is represented by N bits. Table 4 gives the
Here x, y represent the adjacent grey scale pixel values of results of entropy analysis of grey scale test images. It
the test image. N denotes the total number of duplets (x, is observed that the entropy reaches towards the ideal
y) obtained from the test image. For a good cryptosys- value 8. Table 5 compares the information entropy of
tem, the correlation coefficients for the encrypted image our proposed method with the other recently proposed
should be very small, i.e. close to zero. method.
20 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM

Table 4: Information entropy of plain and encrypted images. Table 6: NPCR and UACI values for one pixel difference in a
Images Plain image Encrypted image plain image.
Lena 7.5690 7.9975 Images Pixel positions Changed value of pixel NPCR UACI
Cameraman 7.0097 7.9974 Lena (1,1) 137 → 138 99.63 33.42
Barbara 7.3975 7.9972 (17, 243) 21 → 22 99.62 33.52
(100,200) 25 → 26 99.63 33.56
(130,11) 56 → 55 99.60 33.46
Table 5: Comparative studies of information entropy. (200,210) 73 → 74 99.63 33.39
(256,256) 72 → 73 99.66 33.24
Encryption algorithm Encrypted Lena Encrypted Cameraman
Mean value 99.63 33.43
Ref. [3] 7.989020 –
Ref. [9] 7.9970 7.9967 Cameraman (1,1) 156 → 157 99.60 33.34
Ref. [12] 7.9970 7.9972 (17, 243) 155 → 156 99.60 33.50
Ref. [14] 7.9895 – (100,200) 162 → 163 99.60 33.61
Ref. [20] 7.9974 – (130,11) 180 → 181 99.60 33.52
Ref. [21] 7.9826 7.9637 (200,210) 133 → 134 99.64 33.46
Proposed algorithm 7.9975 7.9974 (256,256) 113 → 114 99.56 33.47
Mean value 99.60 33.48
Barbara (1,1) 85 → 86 99.60 33.58
(17, 243) 10 → 11 99.60 33.41
6.6 Differential Attack Analysis (100,200) 120 → 121 99.60 33.38
(130,11) 184 → 185 99.61 33.50
The differential attack is a technique of finding the rela- (200,210) 27 → 28 99.57 33.45
tion between two encrypted images making a small vari- (256,256) 89 → 88 99.62 33.48
Mean value 99.60 33.47
ation in the plain image. For a good cryptosystem, a
small variation in input image should generate a totally
different encrypted image. Two metrics: Number of Pix- Table 7: Comparison of NPCR and UACI for encrypted Lena.
els Change Rate (NPCR) and Unified Average Changing Encryption algorithm NPCR UACI
Intensity (UACI) are used to analyse the strength of an Ref. [4] 99.61 33.46
Ref. [5] 99.62 27.38
encryption algorithm against differential attack. NPCR Ref. [8] 99.63 33.71
is defined as the percentage of the number of differ- Ref. [11] 99.60 33.44
Ref. [14] 99.65 33.34
ent pixels at the same positions between two encrypted Ref. [15] 99.62 33.06
images whose plane images are of only one pixel differ- Ref. [16] 99.62 33.51
ence. UACI indicates the average intensity of differences Ref. [17] 99.61 33.50
Ref. [20] 99.62 33.51
between two encrypted images with one pixel change in Ref. [21] 99.61 33.33
their plane images. Let us consider two encrypted images Proposed method 99.64 33.56
E1 and E2 and their corresponding plain images differ
only by one pixel. Let E1 (i, j) and E2 (i, j) are the pixel and UACI [22,23] for two true random images may be
values at location (i, j) of E1 and E2 , respectively. Now, expressed by the Equations of (25) and (26). For N = 8,
we construct a bipolar array D with elements 0 or 1 NPCRexpd = 99.609% and UACIexpd = 33.464%.
using equation (22). The dimension of D is same as the
 
dimensions of E1 and E2 . 1
NPCRexpd = 1 − N × 100% (25)
 2
1, if E1 (i, j) = E2 (i, j)
D(i, j) = . (22)
0, Otherwise 2N −1
1 i=1 i(i + 1)
The NPCR and UACI are computed by applying the UACIexpd = × 100%. (26)
22N 2N −1
following formulas (23)–(24).

i,j D(i, j) 6.7 Randomness Test
NPCR = × 100%, (23)
r×c
The verification of randomness of a set of digital numbers
 
i,j |E1 (i, j) − E2 (i, j)|
1 is essential for demanding it as a chaotic sequence. The
UACI = × 100%. (24) National Institute of Standards and Technology (NIST)
r×c 2N − 1
Commission has released the random and pseudo-
NPCR and UACI metrics are computed on various plain random sequence test standard SP 800–22 on April 2010
images and the results are tabulated in Table 6. A com- [24]. This test standard consists of 16 different tests and
parative study of different recent encryption systems has each test will produce a p-value as the result. If the
been displayed in Table 7 based on the values of NPCR p-value ≥ α (In SP 800–22 test α = 0.01) then we can say
and UACI. The theoretically expected values of NPCR that the digital sequence passed the randomness test. A
 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM 21

Table 8: SP 800–22 random and pseudo-random sequence Table 9: Encryption time of various test images.
test. Test image Dimension Encryption time
Test name Chaotic key Cipher image Lena 256 × 256 0.1641 s
Cameraman 256 × 256 0.1745 s
p-Value Remarks p-Value Remarks
Barbara 256 × 256 0.1631 s
Frequency 0.3154 Success 0.3845 Success
Block frequency 0.5113 Success 0.4615 Success
(M = 128)
Runs 0.3523 Success 0.4529 Success 7. CONCLUSION
Longest runs of 0.7632 Success 0.2615 Success
one (M = 128,
N = 49, K = 5)
A novel image encryption algorithm using 6D hyper-
Rank 0.4224 Success 0.3745 Success chaotic Lorenz system and DNA encoding/decoding
Spectral DFT 0.4253 Success 0.7381 Success technique is presented in this paper. The proposed
Non overlapping 0.5527 Success 0.6642 Success
templates encryption system is fast and highly secured. The secu-
(m = 9, rity against different attacks is analysed. It is seen that
B = 101000111)
Overlapping 0.1637 Success 0.3427 Success the ciphered image is purely dependent on the plain
templates image pixel, previous ciphered pixel, chaotic key and
(M = 1032,
N = 508)
DNA encoding/decoding rule. A single bit change in
Universal (L = 6, 0.8932 Success 0.9115 Success the key or the plain image produces an entirely dif-
Q = 640, ferent encrypted image. The time taken for encryption
K = 86741)
Linear complexity 0.7328 Success 0.6917 Success is less than one second which is good enough. The
(M = 500) results obtained from different security tests are also very
Serial (m = 10) p-Value 1 0.8632 Success 0.8825 Success
p-Value 2 0.8315 Success 0.8937 Success good and detailed comparison with other algorithms
Approximate 0.6652 Success 0.7268 Success reported in the literature is also done. The proposed cryp-
entropy
(m = 10) tosystem can be implemented for practical encryption
Cumulative sums Forward 0.3526 Success 0.3647 Success applications.
Reverse 0.3862 Success 0.2973 Success
Random excursions 0.9581 Success 0.8385 Success
(x = 1)
Random excursions 0.7029 Success 0.6572 Success ACKNOWLEDGEMENTS
variant (x = 1) Authors are very thankful to the anonymous reviewers for their
valuable comments and suggestions. Authors also thank the
editor-in-chief for his smooth coordination.
chaotic sequence has good randomness if it passes all the
tests. Table 8 shows the results of SP 800–22 test for the DISCLOSURE STATEMENT
chaotic key and the encrypted image. The total number of
No potential conflict of interest was reported by the
bits of this system are 524288 by considering a 256 × 256 authors.
image in grey scale representation.

REFERENCES
6.8 Speed Analysis 1. L. Hong, B. Ming, and H. Hui, “New image encryption
algorithm based on logistic map and hyper-chaos,” in Pro-
The encryption time is also considered as a perfor-
ceedings of the IEEE International Conference on Com-
mance parameter of a cryptosystem. The cryptosystem putational and Information Sciences, Shiyang China, 2013,
should be as quick as possible in encrypting and decrypt- pp. 713–16.
ing images. Our cryptosystem basically consists of XOR
operations, modular operations and arithmetic opera- 2. L. Min, K. Hu, L. Zhang, and Y. Zhang, “Study on pseudo
tions like multiplication, addition, subtraction to reduce randomness of some pseudorandom number generators
with application,” in Proceedings of the IEEE International
the execution time as far as possible. The algorithm has Conference on Computational Intelligence and Security,
been tested and designed in MATLAB R2009a running Emeishan China, 2013, pp. 569–74.
on an Intel(R) Core(TM) i3–5010U CPU @ 2.10GHz,
4GB memory with Windows 10 Home Single Language 3. Z. Lin, and H. Wang, “Efficient image encryption using a
64-bit operating system. Table 9 lists the encryption times chaos-based PWL memristor,” IETE Tech. Rev., Vol. 27, no.
4, pp. 318–25, Aug. 2010.
taken by our cryptosystem tested on various plain images.
The encryption algorithm is run 10 times for each image 4. C. Fu, W. Li, Z. Meng, T. Wang, and P. Li, “A Symmetric
and the mean execution time is calculated and given in Image Encryption Scheme Using Chaotic Baker map and
Table 9. Lorenz System,” in Proceedings of the IEEE International
22 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM

Conference on Computational Intelligence and Security, 14. X. L. Huang, and G. D. Ye, “An image encryption algorithm
Emeishan China, 2013, pp. 724–8. based on hyper-chaos and DNA sequence,” Multimed.
Tools. Appl., Vol. 72, no. 1, pp. 57–70, Sep. 2014.
5. R. Krishnamoorthi, and P. Murali, “Chaos based image
encryption with orthogonal polynomials model and bit 15. J. Anchal, and R. Navin, “A robust image encryption
shuffling,” in Proceedings of the IEEE International Con- algorithm resistant to attacks using DNA and chaotic logis-
ference on Signal Processing and Integrated Networks, tic maps,” Multimed. Tools. Appl., Vol. 75, no. 10, pp.
Noida India, 2014, pp. 107–12. 5455–72, May 2016.

6. S. Chapaneri, and R. Chapaneri, “Chaos based image 16. J. Chen, Z. Zhu, C. Fu, H. Yu, and L. Zhang, “An efficient
encryption using Latin rectangle scrambling,” in Proceed- image encryption scheme using gray code based permuta-
ings of the Annual IEEE India Conference (INDICON), tion approach,” Opt. Lasers Eng., Vol. 67, pp. 191–204, Apr.
Pune India, 2014, pp. 1–6. 2015.

7. S. Bora, P. Sen, and C. Pradhan, “Novel color image encryp- 17. W. Zhang, H. Yu, Y. Zhao, and Z. Zhu, “Image encryp-
tion technique using Blowfish and Cross Chaos map,” tion based on three-dimensional bit matrix permutation,”
in Proceedings of the IEEE International Conference on Signal. Processing., Vol. 118, pp. 36–50, Jan. 2016.
Communications and Signal Processing, Melmaruvathur
India, 2015, pp. 879–83. 18. Q. Yang, W. M. Osman, and C. Chen, “A New 6D hyper
chaotic system with four positive lyapunov exponents
8. L. Liu, and S. Miao, “A new image encryption algorithm coined,” Intl. Journal. Bifur. Chaos, Vol. 25, no. 4, pp. 1–18,
based on logistic chaotic map with varying parameter,” Apr. 2015.
Springer Plus, Vol. 289, pp. 1–12, Mar. 2016.
19. J. D. Watson, and F. H. C. Crick, “Molecular structure of
9. X. Wang, L. Liu, and Y. Zhang, “A novel chaotic block nucleic acids: A structure for deoxyribose nucleic acid,”
image encryption algorithm based on dynamic random Nature, Vol. 171, pp. 737–8, 1953.
growth technique,” Opt. Lasers Eng., Vol. 66, pp. 10–18,
Mar. 2015. 20. L. Xu, Z. Li, J. Li, and W. Hua, “A novel bit-level image
encryption algorithm based on chaotic maps,” Opt. Lasers
10. M. Kar, M. K. Mandal, and D. Nandi, “RGB image encryp- Eng., Vol. 78, pp. 17–25, Mar. 2016.
tion using hyper chaotic system”, Third International Con-
ference on Research in Computational Intelligence and 21. Y. Luo, M. Du, and J. Liu, “A symmetrical image encryption
Communication Networks (ICRCICN), Kolkata, 2017, pp. scheme in wavelet and time domain,” Commun. Nonlinear
354–9. https://ieeexplore.ieee.org/document/8234534 Sci. Numer. Simulat, Vol. 20, no. 2, pp. 447–60, Feb. 2015.

11. P. Zhen, G. Zhao, L. Min, and X. Jin, “Chaos-based image 22. C. Fu, J. Chen, H. Zou, W. Meng, Y. Zhan, and Y. Yu,
encryption scheme combining DNA coding and entropy,” “A chaos-based digital image encryption scheme with an
Multimed. Tools. Appl., Vol. 75, no. 11, pp. 6303–19, June improved diffusion strategy,” Opt. Express, Vol. 20, no. 3,
2016. pp. 2363, Jan. 2012.

12. X. Wang, Y. Zhang, and X. Bao, “A novel chaotic image 23. M. Kar, M. K. Mandal, D. Nandi, A. Kumar, and S. Banik,
encryption scheme using DNA sequence operations,” Opt. “Bit-plane encrypted image cryptosystem using chaotic
Lasers Eng., Vol. 73, pp. 53–61, October 2015. quadratic and cubic maps,” IETE Tech. Rev., Vol. 33, pp.
651–61, Feb. 2016.
13. R. Enayatifar, H. J. Sadaei, A. H. Abdullah, M. Lee, and I.
F. Isnin, “A novel chaotic based image encryption using 24. NIST, Guide to the Statistical Tests, April 2010. http://nvl
a hybrid model of deoxyribonucleic acid and cellular pubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication
automata,” Opt. Lasers Eng., Vol. 71, pp. 33–41, Aug. 2015. 800-22r1a.pdf
 M. KAR ET AL.: IMAGE ENCRYPTION USING DNA CODING AND HYPERCHAOTIC SYSTEM 23

Authors technology in 2012. His area of research includes computer

security and cryptography, secure chaotic communication,
Madhumita Kar received her BSc degree video coding. He has published more than 15 research papers
in physics (Hons.) from Burdwan Univer- in national and international journals and one patent. He is an
sity, India, in 2005, and the MSc and BEd Associate Professor in the Department of Information Technol-
degrees from the same university in 2007 ogy, National Institute of Technology, Durgapur, India.
and 2011, respectively. She is currently
doing research work in the Department of E-mail: debashisn2@gmail.com
Physics, National Institute of Technology, Mrinal Kanti Mandal received the BSc
Durgapur, as a part time PhD research degree in physics (Hons.) from Burdwan
scholar. Her research interests are on nonlinear dynamics, University, India, in 1998, and the MSc
chaos and cryptography. and PhD degrees from the same univer-
Corresponding author. E-mail: madhumitakar2007@gmail.com sity in 2000 and 2008 respectively. Since
2003 he has been with the Department
Avishek Kumar received BTech degree of Physics, National Institute of Technol-
from department of Electronics and ogy, Durgapur, as an Assistant Professor.
Telecommunication Engineering, Amra- He has published more than 35 papers in
vati University, in 2010. Presently he is national and international journals and 21 conference papers in
pursuing MTech degree from Department proceedings. His research interests include design of electronic
of Physics, National Institute of Tech- circuits & systems, nonlinear dynamics & chaos, cryptography
nology, Durgapur. His area of interest and image processing. He is a reviewer of Nonlinear Dynamics,
includes signal processing, digital image Security and Communication Networks, International Journal of
processing and cryptography. Electronics, Indian Journal of Pure and Applied Physics, Indian
E-mail: kumar.avishek07@gmail.com Journal of Physics, etc. He is a life member of IEEE, IETE, IPS
and IAPT. He was published his bio-data in the science and
Debashis Nandi received his BE degree technology chapter of “Marquis Who’s Who” in 2010 edition.
in electronics and communication engi-
neering from R E College, Durgapur (Uni- E-mail: mrinalkanti.mandal@phy.nitdgp.ac.in
versity of Burdwan), India, in 1994 and
MTech degree from Burdwan University
in microwave engineering in 1997. He
received his PhD degree from IIT Kharag-
pur, India, on medical imaging