Received November 7, 2020, accepted November 30, 2020, date of publication December 8, 2020,

date of current version December 18, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.3043240

An Image Encryption Algorithm Based on

Compressive Sensing and M Sequence
YUQIANG DOU AND MING LI
Big Data Engineering Laboratory for Teaching Resources and Assessment of Education Quality of Henan Province,
Henan Normal University, Xinxiang 453007, China
Corresponding author: Yuqiang Dou ([email protected])
This work was supported in part by the National Natural Science Foundation of China under Grant 62072159 and Grant U1804164, and in
part by the Ph.D. Scientific Research Foundation of Henan Normal University under Grant qd18027.

ABSTRACT In this article, a new image encryption algorithm based on compressive sensing (CS) and
M sequence is proposed to decrease the image communication load and improve the security of image
communication in the internet of things. Most of the available image encryption schemes are based on
chaotic systems to shuffle the image pixels. Before shuffling the image pixels, the random sequence, which
is produced by a chaotic system, need to be sorted. This sorting operation is avoided by utilizing a modified
linear feedback shift register (LFSR) state sequence. Then, the security of the proposed scheme is improved
by combining CS with an improved 1D chaotic system, which is used to construct a measurement matrix.
The computational complexity is reduced by the use of the improved 1D chaotic system. Simultaneously,
the amount of image data is reduced. Simulation results and performance analyses demonstrate that the
proposed encryption scheme can greatly reduce the amount of image data and has good security and
robustness.

INDEX TERMS M sequence, compressive sensing, image encryption, improved 1D chaotic map.

I. INTRODUCTION Compressive sensing (CS) is a novel sampling paradigm.

Image encryption is a critical issue for confidentiality and The concept of CS was formally proposed by Candes and
security. Chaotic maps have been widely used in image Donho In 2006 in [12], [13]. The theory of CS Mainly
encryption owing to the characteristics of high sensitivity to includes three aspects: sparse representation, measurement
initial conditions and control parameters in [1]–[4]. In the matrix and signal reconstruction. Measurement matrix needs
recent years, some other technologies have been combined to satisfy Restricted isometry property (RIP) in [14] or the
with chaotic systems to encrypt images. For example, the sen- robust null space property (RNSP) in [15], [16].As for mea-
sitivity mechanism is built up by utilizing the information surement matrix it can be Gaussian, Bernoulli and Fourier
entropy of the plain-image in [5]. A novel image encryp- in [17], or binary measurement matrix in [15]. As for recon-
tion algorithm is introduced based on the two-dimensional struction algorithms, by using l1 norm minimization the orig-
logistic map and the Latin square image cipher in [6]. inal signal can be exactly recovered by solving a simple
An image encryption scheme combining chaos-memory cel- convex optimization problem through linear programming in
lular automata and weighted histogram is proposed in [7]. [18], [19]. Some reconstruction techniques are based on l0
DNA computing in [8]–[11] is applied to chaos based image norm minimization such as Matching Pursuit (MP) in [20],
encryption system due to its several good characteristics, Orthogonal Matching Pursuit (OMP) in [21] and SL0 algo-
such as massive parallelism, huge storage and ultralow power rithm in [22]. A sparse recovery algorithm by utilizing the
consumption, etc. With the development of the internet of probability-based prior information is proposed in [23] for
things, the amount of image data is constantly increasing compressed sensing system.
and the images needs to be compressed. Hence, compressive Recently, it is necessary to compress and encrypt
sensing becomes a hotspot. plain images simultaneously and many novel encryption
schemes based on CS and chaotic systems have been pro-
The associate editor coordinating the review of this manuscript and posed [24]–[34]. A novel digital image information encryp-
approving it for publication was Lefei Zhang . tion method based on CS and Arnold transformation is

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
220646 VOLUME 8, 2020
Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

proposed in [25]. To further enhance the performance, process can be avoided by the use of the modified LFSR
the double random phase encoding (DRPE) optical encryp- state sequence. Meanwhile, the better permutation effect is
tion technique is utilized. An efficient image compression- attained. Then, CS is utilized to reduce the data amount of
encryption scheme based on Chen’s hyper-chaotic system and the plain image through the measurement matrix, which is
2D compressive sensing is proposed in [26], in which the produced by the improved 1D chaotic system. Furthermore,
hyper-chaotic system is used to generate two measurement the security of the proposed encryption scheme is improved
matrices. A novel encrypted compressive sensing of images by regarding the control parameters of the improved 1D
based on fractional order hyper chaotic Chen system and chaotic system as secret keys. For the meantime, the com-
DNA operation is proposed in [27] in which 4-Dimensioanal putational complexity is reduced by the use of the improved
Fractional order hyper chaotic chen system is used to generate 1D chaotic system.
the measurement matrix. A game-of-life-based scrambling Furthermore, the SHA-512 hash function value of the
method (PDGSM) in [28] is introduced to shuffle the sparse plain image is used to produce the control parameters of the
coefficient matrix of the plain image, in which the respective improved 1D map and linear feedback shift register (LFSR)
permutation matrix is constructed by the rules of game of life. to increase the correlation of the proposed encryption scheme
In this encryption scheme, chaotic sequences are generated and the plain image.
from a 5D memristive hyper chaotic system. This article is organized as follows. In Section II a brief
However, in most of the above algorithms combining com- introduction about compressive sensing is firstly given. Then,
pressive sensing and chaotic maps the measurement matrices M sequence and the improved chaotic map are elaborated.
are generated by adopting multi-dimensional chaotic maps. The proposed image encryption scheme is explained in
Owing to the complex structure and multiple parameters of Section III. In Section IV, the experimental results and the
the multi-dimensional chaotic maps, the difficulty of imple- performance of the proposed scheme are discussed. Finally,
mentation and the computation complexity increase. On the the conclusion is drawn in Section V.
contrary, 1D chaotic maps in [35]–[40] have the advantages
that their structures are simple and they are easy to imple- II. PRELIMINARIES
ment. In this article, an improved 1D chaotic map in [41] A. M SEQUENCE
is used to construct the measurement matrix for compressed The logic expression is of n-order LFSR as the following
sensing.
Linear feedback shift register (LFSR) have been widely an = C1 an−1 ⊕ C2 an−2 ⊕ C3 an−3 ⊕ · · · ⊕ a0 (1)
used in modern communication systems, coding theory and where ⊕ represents module-2 sum. The flow is shown in
cryptography. The mathematical theory of M sequences Fig.1. Each register state will change as the clock triggers.
is firstly described in [42]. A pseudo random n-order M
sequence is the longest linear feedback of a shift regis-
ter sequence of 2n −1 length in which consecutive n-length
strings exactly experience all binary n-length strings in [43].
To improve the linear complexity of LFSR sequences, nonlin-
ear feedback shift register sequences (NLFSR) are presented
such as [44], [45]. However, NLFSR sequences don’t experi-
ence all binary n-length strings only once.
M-sequences have attracted much attention due to
their good randomness properties [46]–[49]. Based on the FIGURE 1. n-order LFSR flow.
knowledge of finite fields, many research results about
M-sequences have been obtained in [47], [48]. As for Move an from the left to the right of Equation 1 and the
constructing M-sequences, some combinatorial algorithms following equation is obtained
are presented in [52]–[54]. Recently, the new construction n
method by the semi-prefer-XOR rule is proposed in [55]. 0=
X
Ci an−i . (2)
In general, M sequence can be used in MIMO radar sys-
i=0
tems in [56], magneto-acoustic imaging in [57] and stream
cipher in [58]. Define a polynomial of F (x) corresponding to Equation (3)
According to the above analyses, a novel image com- as below
n
pression and encryption algorithm is introduced based on X
compressed sensing, M sequence and the improved 1D sys- F (x) = Ci x i (3)
tem. M sequence is one of stream cipher, which is usually i=0
used to change image pixel value by XOR operation. In this which is called characteristic polynomial. If the characteristic
article, the LFSR state sequence is modified into an appropri- polynomial is a primitive one, the pseudo random sequence
ate sequence which experiences all binary n-length numbers generated by the LFSR will have a 2n − 1 length period and
only once. When the plain image is shuffled, the sorting it will be called M sequence.

VOLUME 8, 2020 220647

 Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

The polynomial F (x) of degree n is a primitive polyno- weighting coefficients are zeros or close to zero, s can be
mial if expressed as Equation 6
1. F (x) is irreducible.
s = 9T ξ (6)
2. F (x) is exactly divisible by x m + 1 where m = 2n − 1.
3. F (x) can’t be divisible by x q + 1 where q < m where 9 is the orthogonal basis matrix and ξ is the weighting
The coefficients of F (x) can be treated as a secret coefficient vector in RN . The measurement result y can be
key. The LFSR state sequence can be noted as S = obtained by a measurement matrix 8 with size M × N as the
(an−1 an−2 an−3 · · · a0 ) where the period length of S is 2n − 1. following formula
S can also be treated as a secret key. Convert S into a decimal
y = 8ξ = 89ξ = 2ξ (7)
number of DS by Equation (4).
where N < M and 2 = 89. 2 is called sensor matrix.
DS = an−1 × 2n−1 + an−2 × 2n−2 + an−3 × 2n−3 If the measurement matrix 8 satisfies the Restricted Isom-
+ · · · + a0 (4) etry Property (RIP), the signal s can be recovered by solving
the non-convex problem as the following
DS will experience all the integers from 1 to 2n −1 and each (
value will be experienced only once. The number of n-order min kξ k0
(8)
primitive polynomial is λ = 1n ϕ (2n − 1), where ϕ (·) is s. t. y = 8 = 89ξ = 2ξ
Euler’s totient function.
For example, F (x) = x 4 + x + 1 is a 4-order primitive . This problem can be solved by the algorithms such as
polynomial. We can get the corresponding linear feedback Matching Pursuit (MP) in [20], Orthogonal Matching Pursuit
expression as Equation (5) (OMP) in [21] and SL0 algorithm in [22].

a4 = a1 ⊕ a0 (5) C. IMPROVED 1D CHAOTIC MAP

1D chaotic maps have simple structures, so that they are
The state transition table is shown in Table 1. As can be seen widely used in image encryption such as Logistic, Sine and
from Table 1, DS will experience all the integers from 1 to Chevyshev maps. To improve the chaotic performance of 1D
15 and each value will be experienced only once in each cycle. chaotic maps the new chaotic map is introduced in [36] which
is defined by the following equation
TABLE 1. State transition.  
xn+1 = Fchaos (u, xn ) × 2k − floor Fchaos (u, xn ) × 2k
(9)
where Fchaos (u, xn ) is one of 1D chaotic maps as Logistic
map, Sine map and Chebyshev map. xn is the output chaotic
sequence, which is acquired by iterating the new chaotic sys-
tem M×3N+N0 times and discarding the former N0 elements.
u ∈ (0 , 10] and k ∈ [8, 20] are the control parameters. x0 is
the initial value. The parameters of x0 , u and k are utilized as
the security keys.
• Logistic-Logistic system (LLS)
The logistic map, which is a simple dynamic nonlinear equa-
tion with complex chaotic behavior, is one of the famous 1D
chaotic system. It can be expressed in the following equation
xn+1 = FL (u, xn ) = u × xn × (1 − xn ) (10)
where u is a control parameter with range of u ∈ (0 , 4] x0 is
the initial value of chaotic map, and xn is the output chaotic
sequence. LLS is defined as
xn+1 = u × xn × (1 − xn ) × 2k
 
− floor u × xn × (1 − xn ) × 2k (11)
• Sine-Sine system (SSS)
The Sine map is one of 1D chaotic maps and has a similar
B. COMPRESSIVE SENSING
chaotic behavior with the Logistic map. The definition can
be described by the following equation.
If a signal of s in RN can be sparsely represented by an
orthogonal basis in CS theory [12], which means its most xn+1 = FS (r, xn ) = r × sin (π × xn ) (12)

220648 VOLUME 8, 2020

Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

where parameter r ∈ (0 , 1] and xn is the output chaotic can be converted to 64 decimal numbers: h1 , h2 , · · · , h64 .
sequence. SSS is defined as the following equation The corresponding initial values are computed via
h1 ⊗ h2 ⊗ · · · ⊗ h5

xn+1 = r × sin (π × xn ) × 2k 
 x0 = + t1
  
 256
− floor r × sin (π × xn ) × 2k h6 ⊗ h7 ⊗ · · · ⊗ h10

(13) (16)
u= × 10 + t2
 256
 k = h11 ⊗ h12 ⊗ · · · ⊗ h15 × 3 + 8 + t3

• Chebyshev-Chebyshev system (CCS) 

The Chebyshev map is also one of 1D chaotic systems and 64
can be described by the following equation. t1 , t2 and t3 are the external keys.

xn+1 = FC (a, xn ) = cos (a × arccos xn ) (14) B. GENERATION OF THE INITIAL CONDITION OF LFSR
If LFSR is n-order, there should be the condition of 2n − 1 ≥
where parameter a ∈ N . CCS is defined as
M × N where M × N is the pixel number of the plain image.
xn+1 = cos (a × arccos xn ) × 2k The initial state can be acquired by selecting n bit binary
  values from 512 bit hash values.
− floor cos (a × arccos xn ) × 2k (15)

The dynamical analysis in [39] has shown that the C. PROCESS OF ENCRYPTION ALGORITHM
improved 1D chaotic systems have excellent chaotic behav- As shown in Fig. 4, the steps for the encryption algorithm can
iors. For example, the bifurcation diagram and Lyapunov be descripted as follows.
exponent of LLS are shown in Fig.2 and Fig.3, which indicate Step 1: The SHA-512 hash function of the plain image is
that The chaotic range of LLS is (0, 10) and is much larger generated and the secret keys are obtained according to Part
than Logistic System. A and Part B of section III
Step 2: The sparse coefficients matrix α is obtained by
transform the plain image I into discrete wavelet transform
(DWT) domain of 9.
 Step 3: Convert α into the 1D image pixel matrix P =
p1 , p2 , · · · , pM ×N with length M × N .
Step 4: The n-order LFSR state sequence DS is obtained
by selecting a primitive polynomial of F (x) according to the
initial value. The permutation position sequence T is obtained
by Algorithm 1.

FIGURE 2. The bifurcation diagram of (a)Logistic map; (b)LLS. Algorithm 1 Obtain the Permutation Sequence DS
1: Input: Initial condition
2: set k = 0
3: for i from 1 to 2n −1
4: S (i) = (an−1 an−2 an−3 · · · a0 ) ;
5: DS (i) = an−1 × 2n−1 + an−2 × 2n−2 + an−3
× 2n−3 + · · · + a0
6: if DS(i) <= M ∗ N
7: T(k) = DS(i);
8: k = k+1;
9: end
10: end
FIGURE 3. The Lyapunov exponent of (a)Logistic map; (b)LLS.

 0Step0 5: Obtain the permuted image pixel vector P’ =

p1 , p2 , · · · , p0M ×N by the permutation position vector T
III. PROPOSED IMAGE ENCRYPTION SCHEME according to the following equation.
A. GENERATION OF THE INITIAL CONDITION OF THE P’ (i) = P (T (i)) . (17)
IMPROVED 1D CHAOTIC MAP
SHA–512 hash function is used to generate the control Step 6: Reshape P’ into Matrix U with size M × N. The
parameters of the 1D improved map for establishing the rela- measurement matrix 8 of size M × N is constructed by
tionship between the encryption scheme and the plain image. the improved chaotic sequence X according to [39] with the
A sequence of hexadecimal value that consists of 512 bits is initial condition of (x0 , u, k). The measurement result Y is
generated by the SHA–512 hash function. The 512-bit hash acquired by
value of the plain image is divided into 8-bit blocks which Y = 8U . (18)

VOLUME 8, 2020 220649

 Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

FIGURE 4. Flowchart of the proposed encryption scheme.

FIGURE 5. Flowchart of the decryption scheme.

Step 7: Q is obtained by quantifying the elements of matrix Y Step 1: Produce the chaotic sequence X according to [36]
into the range from 0 to 255 with the sigmoid function in [60]. with the secret keys and obtain W according to X . Matrix Q
Step 8: The diffusion matrix W is obtained by the following is obtained by executing the inverse diffusion process.
equation Step 2: Y is obtained by executing the inverse quantization
   according to the inverse sigmoid function.
W (i) = mod floor X (i) × 1014 , 256 . (19) Step 3: Construct the measurement matrix 8 and recon-
Step 9: Reshape matrix Q into a sequence of Q’ and obtain struct the matrix U from Y by SL0 algorithm.
the diffused sequence C by the following equations. Step 4: Obtain the LFSR state sequence DS and acquire the
( permutation sequence T according to Algorithm 1.
C (1) = mod (C (0) + Q’ (1) + W (1) , 256) Step 5: Obtain the sparse coefficient matrix α and
(20)
C (i) = mod (C (i − 1) + Q’ (i) + W (i) , 256) acquire the original plain image by inverse discrete wavelet
transform (IDWT).
Step 10 Reshape sequence C into matrix C’ with the same
size of Y and C’ is the final encrypted image.
IV. SIMULATIONS AND SECURITY ANALYSIS
D. PROCESS OF THE DECRYPTION ALGORITHM Simulation results and security analyses of the proposed
The decryption procedure is the reverse of encryption pro- encryption scheme are given out in this section. Eight gray-
cess. The flowchart of the decryption is shown in Fig.5. The scale images, including ‘Lena’, ‘Pepper’, ‘Baboon’, ‘Boat’,
detailed process is as follows. ‘man’, ‘Lake’, ‘Run’ and ‘Sat’, which are shown in Fig. 6,

220650 VOLUME 8, 2020

Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

FIGURE 6. Eight test images.

FIGURE 7. Cipher images (a1)-(c1) of Lena when CR = 0.25, 0.5 and 0.75; FIGURE 8. Cipher images (a1)-(c1) of Pepper when CR = 0.25, 0.5 and
Decrypted images: (a2)-(c2) when CR = 0.25, 0.5 and 0.75; 0.75; Decrypted images: (a2)-(c2) when CR = 0.25, 0.5 and 0.75;

with size 256 × 256 are chosen to validate the proposed volume ratio between compressed image and plain image
scheme in the following experiments. Simulation exper- which is computed according to the following equation
iments have been carried out on an Intel(R) Core(TM)
i7-9700 processor with 3 GHz and 32 GB memory capac- hc × wc
CR = (21)
ity. The simulation software is MALTAB R2016b. The ini- hI × wI
tial parameters of the proposed encryption algorithm are
set as x0 = 0.5321, u1 = 5.4374, t0 = 0.4560, and where hI and wI denote the height and width of the original
k1 = 14. The initial state of LFSR is set a binary number image and hc and wc are that of the cipher image.
of ‘11111111111111111’. The primitive polynomial coeffi- According to Figs. 7-11, it is evident that the cipher
cients of F (x) is set ‘00100000000000001’. images are unrecognizable and the image data amount is
reduced according to CR. Compared to the original images,
A. ENCRYPTION RESULTS AND DECRYPTION RESULTS the decrypted images have preserved the main information of
FOR DIFFERENT IMAGES the original images and they become more similar when CR
Five plain gray images are utilized as the test images includ- is increasing.
ing ‘Lena’, ‘Pepper’, ‘Baboon’, ‘Run’ and ‘Sat’.
The encrypted and corresponding decrypted results are B. PSNR AND SSIM ANALYSIS
illustrated in Figs. 7-11 where compression ratio is changing The quality of the reconstructed image is evaluated by the
from 0.25 to 0.75. Compression ratio (CR) is the image Peak Signal to Noise Ratio (PSNR) which is calculated by

VOLUME 8, 2020 220651

 Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

TABLE 2. Comparison of PSNR of reconstructed image LENA for different measuring matrix.

FIGURE 9. Cipher images (a1)-(c1) of Baboon when CR = 0.25, 0.5 FIGURE 11. Cipher images (a1)-(c1) of Baboon when CR = 0.25, 0.5 and
and 0.75; Decrypted images: (a2)-(c2) when CR = 0.25, 0.5 and 0.75; 0.75; Decrypted images: (a2)-(c2) when CR = 0.25, 0.5 and 0.75;

in Table 2 with the different CR values. As can be seen

from Table 2, the quality of the reconstructed image of
the proposed scheme is better than the encryption schemes
in [29], [30], [34], [35].
The Structural similarity (SSIM) is used to evaluate the
similarity of two images which is defined by

(2µX µX 0 + C1 )(2σXX 0 + C 2 )
SSIM = (23)
(µ2X + µ2X 0 + C1 )(σX2 + σX2 0 + C2 )

where µX and µX 0 represent the average values of the plain

image X and the reconstructed image X ’. σX and σX 0 are
the corresponding standard deviations. σXX 0 is the cross-
FIGURE 10. Cipher images (a1)-(c1) of Baboon when CR = 0.25, 0.5 and covariance of X and X ’. C1 and C2 are constants. The similar-
0.75; Decrypted images: (a2)-(c2) when CR = 0.25, 0.5 and 0.75; ity between X and X ’ is higher when the SSIM value is higher.
The images of ‘Baboon’, ‘Boat’, ‘Lake’, ‘man’, ‘Peppers’,
‘Run’ and ‘Sat’ are utilized to verify our proposed scheme.
the following formula The SSIM values of our proposed scheme are higher than the
2552 schemes in [29], [30], [34], [35] according to Table 3.
PSNR = 10 log10 ( ) (22)
M P
N
1
(i, j) − X 0 (i, j)]2
P
MN [X C. HISTOGRAM ANALYSIS
i=1 j=1
Histogram graphically represents the distribution of the pixel
where X (i, j) is the plain image and X 0 (i, j) is the recon- values of an image. It is needed that the histogram of the
structed image. The image of ‘Lena’ is used to verify our encrypted image has a uniform distribution. Fig. 12 shows
proposed encryption scheme. The PSNR values are listed that the histograms of the plain images and the histograms

220652 VOLUME 8, 2020

Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

TABLE 3. SSIM of different algorithms.

FIGURE 12. a1-d1: histograms of four plain images; a2-d2: histograms of corresponding cipher images.

of their corresponding encrypted images are uniform. Hence, corresponding cipher image with CR equal to 0.5. According
the proposed scheme is robust against statistical attacks. to Table 4 the correlation coefficients of plain image are
near 1 and those of the cipher images are almost equal to 0
D. PIXEL CORRELATION ANALYSIS which means the correlation between the adjacent pixels has
It is well-known that there are high correlations between the been greatly reduced.
adjacent pixels, and a secure encryption algorithm should
reduce the correlation to improve the ability of resistance
E. KEY SPACE ANALYSIS
to statistical attacks. The correlation coefficient is defined
according to the following equation The key space of an image encryption scheme, which can
resist brute-force attacks, should be more than 2100 in [59].
PM
(xi − E (x))(yi − E (y)) In this article the secret keys of the proposed encryption
Cor (x, y) = qP i=1 (24) algorithm mainly include: x0 ∈ (0 , 1], u1 ∈ (0 , 10], k1 ∈
M 2 PM
i=1 (xi −E (x)) i=1 (yi − E (y))2
[8, 20] , S and the coefficient of F (x). S is generally larger
than 216 . If the computational precision of the computer is
where E (x) = M1 M j=1 xj , E (y) = M
1 PM
P
j=1 yj . x and y are 10−14 , the key space is about 1014 × 1014 × 10 × 216 ×
gray values of the adjacent pixels in the images and M is the λ ≈ λ2112 .
total number randomly chosen from the test image. Hence, the key space of our method is large enough to resist
The correlation diagram among adjacent pixels at horizon- all kinds of brute-force attacks.
tal, vertical and diagonal directions of the plain image of
‘Lena’ and the corresponding cipher image with CR equal
to 0.5 are shown as in Fig.13. The size of cipher images are F. KEY SENSITIVITY ANALYSIS
128 × 256. A good encryption scheme should be very sensitive to
Table 4 shows the correlation coefficients of horizontally, secret keys. The sensitivity can be evaluated by Number
vertically and diagonally of the plain images of ‘Baboon’, of pixel change rate (NPCR) and unified average change
‘Boat’, ‘Lake’, ‘man’, ‘Peppers’, ‘Run’ and ‘Sat’ and the intensity (UACI) which are defined by the

VOLUME 8, 2020 220653

 Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

TABLE 4. Comparison of correlation coefficient with other schemes.

following formulas. The image of ‘Lena’ is used to verify key sensitiv-


M X
N
 ity. After slightly changing the value of parameter x0
1 X from 0.5321 to0.53210000000001 the obtained NPCR
NPCR =  D(i, j) × 100% (25)
M ×N and UACI betweenthe cipher images are 99.6570 and
i=1 j=1
  33.3159. After slightly changing the value of parameter x0
M X
N from 0.5321 to 0.53209999999999, the obtained NPCR
1 X |c1 (i, j) − c 2 (i, j)|
UACI =   × 100% and UACI between the cipher images are 99.6151 and
M ×N 255 33.2703. After slightly changing the value of parame-
i=1 j=1
(26) ter u1 from to 5.43740000000001 the obtained NPCR
and UACI between the cipher images are 99.5789 and
where 33.3727. After slightly changing the value of parameter u1
(
0 if c1 (i, j) = c2 (i, j) from 5.43740000000000 to 5.43739999999999 the obtained
D(i, j) = NPCR and UACI between the cipher images are 99.5903 and
1 if c1 (i, j) 6 = c2 (i, j)
33.3532. The results show that the proposed encryption
and c1 and c2 are the encrypted images corresponding to two scheme is highly sensitive to a slight change of the secret
slightly deferent keys. keys.

220654 VOLUME 8, 2020

Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

TABLE 5. Comparison of entropy with other schemes.

FIGURE 13. Correlation analyses of image Lena. a. horizontal correlation of original and encrypted images; vertical
correlation of original and encrypted images; diagonal correlation of original and encrypted images.

G. INFORMATION ENTROPY ANALYSIS It is evident that our encryption algorithm has better perfor-
The information entropy formula is as the following mance than the schemes in. [24], [34], [35].
v
X
H (x) = p(xi ) log 2 (p(xi )) (27) H. ROBUSTNESS ANALYSIS
i=1
Cipher images are easily influenced by noise and data loss
where p(xi ) is the probability of a message of xi and v is the during image transmission. A good image encryption scheme
number of the messages of the information source. For an should be able to resist these attacks. The image of ‘milk-
image with 256 gray levels, v is equal to 256. Table 5 presents drop’ is used to test the ability of the proposed encryption
the information entropies of the plain images and cipher scheme. At first, the image is encrypted by the proposed
images (CR = 0.75). The last row gives out the average scheme and the encrypted image is shown as Fig.14 (a1)
values of the information entropies of the plain images and Then, Fig.14 (a1) is attacked by ‘salt&pepper’ noise whose
cipher images. The information entropies of the correspond- intensity is 0.0001 and the result is shown as Fig.14 (b1).
ing cipher images are all more than 7.99 as shown in Table 5. At last Fig.14 (a1) is cropped with data loss size of 64 × 64

VOLUME 8, 2020 220655

 Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

FIGURE 14. Noise and data loss attack (a1) the encrypted image; (b1) the encrypted image added with noise; (c1) the
encrypted image with 64∗64 loss; (a2-c2) the decrypted image corresponding to (a1-c1).

the cropped encrypted image is shown as Fig.14 (c1). [5] G. Ye, C. Pan, X. Huang, Z. Zhao, and J. He, ‘‘A chaotic image encryption
The corresponding decrypted images are shown as algorithm based on information entropy,’’ Int. J. Bifurcation Chaos, vol. 28,
no. 1, Jan. 2018, Art. no. 1850010.
Fig.14 (a2, b2, c2). As can be seen, they contain the main [6] M. Machkour, A. Saaidi, and M. L. Benmaati , ‘‘A novel image encryption
information of the original image when noise and crop attacks algorithm based on the two-dimensional logistic map and the Latin square
occur. image cipher,’’ 3D Res., vol. 6, Oct. 2015, Art. no. 36.
[7] A. Souyah and K. M. Faraoun, ‘‘An image encryption scheme combining
chaos-memory cellular automata and weighted histogram,’’ Nonlinear
V. CONCLUSION Dyn., vol. 86, no. 1, pp. 639–653, Oct. 2016.
In this article the modified LFSR state sequence is utilized [8] A. Jain and N. Rajpal, ‘‘A robust image encryption algorithm resistant to
to shuffle the sparse coefficient matrix of the plain image and attacks using DNA and chaotic logistic maps,’’ Multimedia Tools Appl.,
vol. 75, no. 10, pp. 5455–5472, May 2016.
good permutation effect has been achieved. Then, the amount
[9] K. Zhan, D. Wei, J. Shi, and J. Yu, ‘‘Cross-utilizing hyperchaotic and
of image data is greatly reduced by the use of CS. For the DNA sequences for image encryption,’’ J. Electron. Imag., vol. 26, no. 1,
meantime, CS is combined with the improved 1D chaotic Feb. 2017, Art. no. 013021.
system, by which the security of the proposed encryption [10] X. Huang and G. Ye, ‘‘An image encryption algorithm based on hyper-
chaos and DNA sequence,’’ Multimedia Tools Appl., vol. 72, no. 1,
scheme is improved. The improved 1D chaotic is used to pp. 57–70, Sep. 2014.
construct the measurement matrix and the initial parameters [11] A. U. Rehman, X. Liao, A. Kulsoom, and S. A. Abbas, ‘‘Selective encryp-
are used as the secret keys. Simulation results demonstrate tion for gray images based on chaos and DNA complementary rules,’’
Multimedia Tools Appl., vol. 74, no. 13, pp. 4655–4677, Jul. 2015.
the proposed encryption scheme can reduce the amount of
[12] D. L. Donoho, ‘‘Compressed sensing,’’ IEEE Trans. Inf. Theory, vol. 52,
image data and improve the security of image communication no. 4, pp. 1289–1306, Apr. 2006.
simultaneously. Performance analyses demonstrate that the [13] R. Baraniuk, ‘‘Compressive sensing [lecture notes],’’ IEEE Signal Process
proposed encryption scheme has a large key space and high Mag, vol. 24, no. 4, pp. 118–121, Jul. 2007.
[14] E. J. Candes and T. Tao, ‘‘Near-optimal signal recovery from random
key sensitivity and good robustness to noise and data loss. projections: Universal encoding strategies?’’ IEEE Trans. Inf. Theory,
For the meantime, our scheme can resist statistical attack and vol. 52, no. 12, pp. 5406–5425, Dec. 2006.
differential attack. [15] M. Lotfi and M. Vidyasagar, ‘‘A fast noniterative algorithm for compres-
sive sensing using binary measurement matrices,’’ IEEE Trans. Signal
Process., vol. 66, no. 15, pp. 4079–4089, Aug. 2018.
REFERENCES
[16] M. Lotfi and M. Vidyasagar, ‘‘Compressed sensing using binary matrices
[1] Y. Zhang, D. Xiao, Y. Shu, and J. Li, ‘‘A novel image encryption scheme of nearly optimal dimensions,’’ IEEE Trans. Signal Process., vol. 88,
based on a linear hyperbolic chaotic system of partial differential equa- pp. 3008–3021, 2020.
tions,’’ Signal Process., Image Commun., vol. 28, no. 3, pp. 292–300, [17] E. J. Candès, ‘‘The restricted isometry property and its implications for
Mar. 2013. compressed sensing,’’ Comp. Rendus Mathematique, vol. 346, nos. 9–10,
[2] Y. Zhang and D. Xiao, ‘‘Double optical image encryption using discrete pp. 589–592, May 2008.
Chirikov standard map and chaos-based fractional random transform,’’ [18] S. Qaisar, R. M. Bilal, W. Iqbal, M. Naureen, and S. Lee, ‘‘Compressive
Opt. Lasers Eng., vol. 51, no. 4, pp. 472–480, Apr. 2013. sensing: From theory to applications, a survey,’’ J. Commun. Netw., vol. 15,
[3] G. Ye and K.-W. Wong, ‘‘An efficient chaotic image encryption algorithm no. 5, pp. 443–456, Oct. 2013.
based on a generalized Arnold map,’’ Nonlinear Dyn., vol. 69, no. 4, [19] E. J. Candes and T. Tao, ‘‘Decoding by linear programming,’’ IEEE Trans.
pp. 2079–2087, Sep. 2012. Inf. Theory, vol. 51, no. 12, pp. 4203–4215, Dec. 2005.
[4] O. Mirzaei, M. Yaghoobi, and H. Irani, ‘‘A new image encryption method: [20] S. G. Mallat and Z. Zhang, ‘‘Matching pursuits with time-frequency dic-
Parallel sub-image encryption with hyper chaos,’’ Nonlinear Dyn., vol. 67, tionaries,’’ IEEE Trans. Signal Process., vol. 41, no. 12, pp. 3397–3415,
no. 1, pp. 557–566, Jan. 2012. Dec. 1993.

220656 VOLUME 8, 2020

Y. Dou, M. Li: Image Encryption Algorithm Based on Compressive Sensing and M Sequence

[21] Y. C. Pati, R. Rezaiifar, and P. S. Krishnaprasad, ‘‘Orthogonal matching [45] A. Kholosha, ‘‘Clock-controlled shift registers and generalized Geffe key-
pursuit: Recursive function approximation with applications to wavelet stream generator,’’ in Proc. Int. Conf. Prog. Cryptol.-Indocrypt (DBLP),
decomposition,’’ in Proc. 27th Asilomar Conf. Signals, Syst. Comput., 2002, pp. 287–296.
Nov. 1993, pp. 40–44. [46] C. Li, X. Zeng, T. Helleseth, C. Li, and L. Hu, ‘‘The properties of a class of
[22] H. Mohimani, M. Babaie-Zadeh, and C. Jutten, ‘‘A fast approach for over linear FSRs and their applications to the construction of nonlinear FSRs,’’
complete sparse decomposition based on smoothed `0 norm,’’ IEEE Trans. IEEE Trans. Inf. Theory, vol. 60, no. 5, pp. 3052–3061, May 2014.
Signal Process., vol. 57, no. 1, pp. 289–301, Jan. 2009. [47] T. Helleseth, ‘‘Nonlinear shift registers—A survey and challenges,’’ in
[23] Q. Jiang, S. Li, Z. Zhu, H. Bai, X. He, and R. C. de Lamare, ‘‘Design Algebraic Curves and Finite Fields: Cryptography and Other Applications.
of compressed sensing system with probability-based prior information,’’ Berlin, Germany: De Gruyter, 2014, pp. 121–144.
IEEE Trans. Multimedia, vol. 22, no. 3, pp. 594–609, Mar. 2020. [48] C. Li, X. Zeng, C. Li, T. Helleseth, and M. Li, ‘‘Construction of de Bruijn
[24] J. Chen, Y. Zhang, L. Qi, C. Fu, and L. Xu, ‘‘Exploiting chaos-based sequences from LFSRs with reducible characteristic polynomials,’’ IEEE
compressed sensing and cryptographic algorithm for image encryption and Trans. Inf. Theory, vol. 62, no. 1, pp. 610–624, Jan. 2016.
compression,’’ Opt. Laser Technol., vol. 99, pp. 238–248, Feb. 2018. [49] X. Wang and L. Jiang, ‘‘Cycle structure and adjacency graphs of a class
[25] X. Liu, Y. Cao, P. Lu, X. Lu, and Y. Li, ‘‘Optical image encryption of LFSRs and a new family of de bruijn cycles,’’ IEEE Access, vol. 6,
technique based on compressed sensing and arnold transformation,’’ Optik, pp. 38770–38779, 2018.
vol. 124, no. 24, pp. 6590–6593, Dec. 2013. [50] S. W. Golomb, Shift Register Sequences. Sydney, NSW, Australia:
[26] N. Zhou, S. Pan, S. Cheng, and Z. Zhou, ‘‘Image compression–encryption Holden-Day, 1982.
scheme based on hyper-chaotic system and 2D compressive sensing,’’ Opt. [51] S. W. Golomb and G. Gong, Signal Design for Good Correlation: For
Laser Technol., vol. 82, pp. 121–133, Aug. 2016. Wireless Communication, Cryptography, and Radar, vol. 18. Cambridge,
[27] S. Kayalvizhi and S. Malarvizhi, ‘‘A novel encrypted compressive sensing U.K.: Cambridge Univ. Press, 2005.
of images based on fractional order hyper chaotic chen system and DNA [52] M. Alhakim, ‘‘A simple combinatorial algorithm for de Bruijn sequences,’’
operations,’’ Multimedia Tools Appl., vol. 79, nos. 5–6, pp. 3957–3974, Amer. Math. Monthly, vol. 117, no. 8, pp. 728–732, 2010.
Feb. 2020. [53] A. Alhakim, ‘‘Spans of preference functions for de Bruijn sequences,’’
[28] Z. Gan, X. Chai, J. Zhang, Y. Zhang, and Y. Chen, ‘‘An effective Discrete Appl. Math., vol. 160, nos. 7–8, pp. 992–998, May 2012.
image compression–encryption scheme based on compressive sensing [54] H. Fredricksen, ‘‘A survey of full length nonlinear shift register cycle
(CS) and game of life (GOL),’’ Neural Comput. Appl., vol. 32, no. 17, algorithms,’’ SIAM Rev., vol. 24, no. 2, pp. 195–221, Apr. 1982.
pp. 14113–14141, Sep. 2020. [55] X. Wang, L. Zhang, and L. Jiang, ‘‘A new effective shift rule for
[29] G. Hu, D. Xiao, Y. Wang, and T. Xiang, ‘‘An image coding scheme using M-Sequences,’’ IEEE Access, vol. 8, pp. 74957–74964, 2020.
parallel compressive sensing for simultaneous compression-encryption [56] M. M. Feraidooni, D. Gharavian, S. Imani, and M. Alaee-Kerahroodi,
applications,’’ J. Vis. Commun. Image Represent., vol. 44, pp. 116–127, ‘‘Designing M-ary sequences and space-time receive filter for moving
Apr. 2017. target in cognitive MIMO radar systems,’’ Signal Process., vol. 174,
[30] X. Chai, X. Zheng, Z. Gan, D. Han, and Y. Chen, ‘‘An image encryption Sep. 2020, Art. no. 107620.
algorithm based on chaotic system and compressive sensing,’’ Signal [57] S. Zhang, R. Ma, T. Yin, and Z. Liu, ‘‘M-sequence-coded excitation for
Process., vol. 148, pp. 124–144, Jul. 2018. magneto-acoustic imaging,’’ Med. Biol. Eng. Comput., vol. 57, no. 5,
[31] X. Chai, X. Fu, Z. Gan, Y. Zhang, Y. Lu, and Y. Chen, ‘‘An efficient chaos- pp. 1059–1067, May 2019.
based image compression and encryption scheme using block compressive [58] L. Jiao, Y. Hao, and D. Feng, ‘‘Stream cipher designs: A review,’’ Sci.
sensing and elementary cellular automata,’’ Neural Comput. Appl., vol. 32, China Inf. Sci., vol. 63, no. 3, pp. 80–104, Mar. 2020.
no. 9, pp. 4961–4988, May 2020. [59] B. Norouzi, S. M. Seyedzadeh, S. Mirzakuchaki, and M. R. Mosavi,
[32] W. Shao, M. Cheng, C. Luo, L. Deng, M. Zhang, S. Fu, M. Tang, ‘‘A novel image encryption based on row-column, masking and main
and D. Liu, ‘‘An image encryption scheme based on hybrid electro- diffusion processes with hyper chaos,’’ Multimedia Tools Appl., vol. 74,
optic chaotic sources and compressive sensing,’’ IEEE Access, vol. 7, no. 3, pp. 781–811, Feb. 2015.
pp. 156582–156591, 2019. [60] J. Han and M. Claudio, ‘‘The influence of the sigmoid function param-
[33] K. Zhu and J. Cheng, ‘‘Color image encryption via compressive sens- eters on the speed of backpropagation learning,’’ in Proc. Int. Workshop
ing and chaotic systems,’’ in Proc. MATEC Web Conf., vol. 309, 2020, Artif. Neural Netw., Natural Artif. Neural Comput. Berlin, Germany:
p. 03017. Springer-Verlag, 1995, pp. 195–201.
[34] N. Zhou, A. Zhang, J. Wu, D. Pei, and Y. Yang, ‘‘Novel hybrid image
compression-encryption algorithm based on compressive sensing,’’ Optik-
Int. J. Light Electron Opt., vol. 125, no. 18, pp. 508–5075, 2014. YUQIANG DOU received the B.E. degree from
[35] R. Ponuma and R. Amutha, ‘‘Compressive sensing based image Zhengzhou University in 2003, the master’s degree
compression-encryption using novel 1D-chaotic map,’’ Multimedia Tools from the Communication University of China
Appl., vol. 77, no. 15, pp. 19209–19234, Aug. 2018. in 2006, and the Ph.D. degree from the Key Lab-
[36] R. R. Kumar and M. B. Kumar, ‘‘A new chaotic image encryption using
oratory of Media Audio and Video, Communica-
parametric switching based permutation and diffusion,’’ ICTACT J. Image
Video Process., vol. 4, no. 4, pp. 795–804, Mar. 2014.
tion University of China, Ministry of Education,
[37] Y. Zhou, L. Bao, and C. L. P. Chen, ‘‘A new 1D chaotic system for image in 2016.
encryption,’’ Signal Process., vol. 97, pp. 172–182, Apr. 2014. He is currently a Teaching Staff Member with
[38] C. Lv-Chen, L. Yu-Ling, Q. Sen-Hui, and L. Jun-Xiu, ‘‘A perturbation the School of Computer and Information Engi-
method to the tent map based on Lyapunov exponent and its application,’’ neering, Henan Normal University. His research
Chin. Phys. B, vol. 24, no. 10, pp. 78–85, 2015. interests include image information hiding, multimedia security, and audio
[39] C. Pak and L. Huang, ‘‘A new color image encryption using combination signal processing.
of the 1D chaotic map,’’ Signal Process., vol. 138, pp. 129–137, Sep. 2017.
[40] C. Pak, K. An, P. Jang, J. Kim, and S. Kim, ‘‘A novel bit-level color image
encryption using improved 1D chaotic map,’’ Multimedia Tools Appl., MING LI received the master’s degree in science
vol. 78, no. 9, pp. 12027–12042, May 2019. from the College of Physics and Information Engi-
[41] L. Yu, J. P. Barbot, G. Zheng, and H. Sun, ‘‘Compressive sensing with neering, Henan Normal University, Henan, China,
chaotic sequence,’’ IEEE Signal Process. Lett., vol. 17, no. 8, pp. 731–734, in 2010, and the Ph.D. degree from the College
Aug. 2010.
of Computer Science of Chongqing University,
[42] N. Zierler, ‘‘Linear recurring sequences,’’ J. Soc. Ind. Appl. Math., vol. 7,
no. 1, pp. 31–48, Mar. 1959. Chongqing, China, in 2014.
[43] J. Mykkeltveit, M.-K. Siu, and P. Tong, ‘‘On the cycle structure of He is currently an Associate Professor with the
some nonlinear shift register sequences,’’ Inf. Control, vol. 43, no. 2, College of Computer and Information Engineer-
pp. 202–215, Nov. 1979. ing, Henan Normal University, China. His research
[44] M. Hell, T. Johansson, and W. Meier, ‘‘Grain: A stream cipher for con- interests include multimedia security, information
strained environments,’’ Int. J. Wireless Mobile Comput., vol. 2, no. 1, hiding, and compressive sensing.
pp. 86–93, 2007.

VOLUME 8, 2020 220657