See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220405355

An Improved Image Encryption Algorithm based on Chaotic System

Article  in  Journal of Computers · November 2009

DOI: 10.4304/jcp.4.11.1091-1100 · Source: DBLP

CITATIONS READS

55 686

3 authors, including:

Liu Shubo Zhengquan Xu

Wuhan University Wuhan University
14 PUBLICATIONS   131 CITATIONS    119 PUBLICATIONS   782 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Zhengquan Xu on 05 June 2014.

The user has requested enhancement of the downloaded file.

JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009 1091

An Improved Image Encryption Algorithm based

on Chaotic System
Shubo Liu1,2, Jing Sun1,2, Zhengquan Xu1
1 State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University,
Wuhan, Hubei, China, 430079
2 Computer School of Wuhan University, Wuhan, Hubei, China, 430079
E-mail:[email protected], [email protected], [email protected]

Abstract—The security of stream cipher, which is known nature will be deteriorated when a finite precision with
as one of the main cipher techniques, is dependents fixed-point arithmetic is used. Moreover, some severe
completely on the quality of generated pseudo-stochastic problems have been observed，such as short cycle length,
sequences. Chaotic systems can produce the
non-ideal distribution and high-correlation[4].
pseudo-random sequences with good randomness ，
In this paper, the protection of images is particularly
therefore, these systems are suitable to the stream cipher.
In this paper, a new encryption algorithm is proposed by
in interested, while most conventional ciphers, such as
analyzing the principle of the chaos encryption algorithm DES, IDEA, and AES [5, 6], are not suitable for image
based on logistic map. Moreover, the security and encryption in real time，because their speed is slow due
performance of the proposed algorithm is also estimated. to a large data volume and strong correlation among
The experimental results based on coupled chaotic maps image pixels. In recent years, a number of chaos-based
approve the effectiveness of the proposed method, and the cryptographic schemes have been proposed [3, 25, 7-22],
coupled chaotic maps shows advantages of large key space where most of them are based on block transform
and high-level security. The ciphertext generated by this
utilizing the chaotic map. To enhance security, the
method is the same size as the plaintext and is suitable for
practical use in the secure transmission of confidential dimension of these maps has been extended from one to
information over the Internet. three. Among them, chaos based encryption techniques
are considered good for practical use as these techniques
Index Terms—Logistic map; Stream cipher; Image provide a good combination of speed, high security,
encryption; complexity, reasonable computational overheads and
computational power etc.
The characteristics of the chaotic maps have attracted
I. INTRODUCTION the attention of cryptographers to develop new
encryption algorithms with higher security and speed.
Chaos theory has been established since 1970s by There are some chaos-based image cryptosystems with
many different research areas, such as physics, their own structure. For example, Pareek et al. [13]
mathematics, engineering, and biology, etc. [1]. Since proposed an image encryption scheme which utilizes two
1990s, many researchers have noticed that there exists chaotic logistic maps and an external key of 80-bit. The
the close relationship between chaos and cryptography [2, initial conditions for both logistic maps were derived
3]. The distinct properties of chaos, such as ergodicity, using the external secret key. The first logistic map was
quasi-randomness, sensitivity dependence on initial used to generate numbers in the range between 1 and 24
conditions and system parameters, have granted chaotic and the initial condition of the second logistic map was
dynamics as a promising alternative for the conventional modified by the numbers generated by the first logistic
cryptographic algorithms. Chaos-based cryptography is map. The authors showed that by modifying the initial
relied on the complex dynamics of nonlinear systems or condition of the second logistic map in this way, its
maps which are deterministic but simple. Therefore, it dynamics became more random. Kwok et al. [15]
can provide a fast and secure means for data protection, proposed a fast chaos-based image cryptosystem with the
which is crucial for multimedia data transmission over architecture of a stream cipher. In particular, the
fast communication channels, such as the broadband plain-image pixels are masked by a pseudo-random
internet communication. keystream generated by a cascade of the skewed tent
Chaos seems to be a good candidate due to its map and a high-dimensional cat map. Behnia et al. [20]
ergodicity and complex dynamics. However, chaotic proposed a novel algorithm for image encryption based
systems are usually assumed to work in the real number on mixture of chaotic maps, using one dimensional
domain, and hence the speed is limited for actual chaotic map and their coupling to obtain high level
implementation. On the other hand, its randomness security [23,24].
It can be noticed that most of the image encryption
Project number：No. 2006CB303104 and No. 40871200. designs are in the form of block cipher, which is usually
Corresponding author: Zhengquan Xu. considered faster than its counterpart, stream cipher,
although stream cipher may provide better security under

© 2009 ACADEMY PUBLISHER

1092 JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009

the concept of perfect security. In this paper, a novel The organization of this paper is as follow. In Section
image encryption algorithm based on logistic map is 2, the design of the proposed chaos-based image
proposed, and it is demonstrated that a well-designed encryption scheme is discussed in detail, while the
chaos-based stream cipher can be a good candidate and design of a new chaos-based pseudo-random generator is
may even outperform the block cipher, on speed and given. The performances and the cryptanalysis of the
security. In it, the keystream generator is based on proposed image encryption scheme are studied in Section
coupled chaotic logistic maps that one logistic chaotic 3. Finally, conclusion remarks are drawn in Section 4.
system generates the satisfied random number to update
the parameter of the other. The chaotic binary sequence II. THE PROPOSED IMAGE ENCRYPTION ALGORITHM
is perturbed by XOR operation on its own three parts.
The encryption step proposed in the algorithm consists of A. Chaos-Based Pseudo-Random Keystream Generator
a simple bitwise XOR operation of the plaintext binary In this paper, the major part of the design is a newly
sequence with the keystream binary sequence to produce proposed chaos-based pseudo-random keystream
the ciphertext binary sequence. Then, a detailed generator (PRKG) based on a couple of chaotic systems.
statistical analysis on the proposed encryption scheme is The structure of the PRKG system is presented in Fig. 1.
given. The experimental results based on coupled chaotic The two logistic chaotic systems use the same principle
maps approve the effectiveness of the proposed method, with different initial values. In our design, the first
and the coupled chaotic maps shows advantages of large logistic chaotic system generates the random numbers to
key space and high-level security. Having a high update the parameters of the second, while some
throughput, the proposed system is ready to be applied in conditions are satisfied. The generator system is
fast real time encryption applications. proposed in the following.

Figure 1. The key stream generator system.

(1)
(a) Quantification Method xn +1 = µ xn (1 − xn )
Generating a pseudorandom binary sequence from the (3)
(2)
yn +1 = µ yn (1 − yn ), n = 0, 1, 2...
orbit of the logistic map
xn +1 = µ xn (1 − xn ) (1) for χn∈(0,1) and µ∈(3.569945,4], evolve successive
states from the first logistic map by χn+1=µ(1) χn(1-χn), and
For χn∈(0,1) and µ∈(3.569945,4], µ and χn are the obtain the real number χn+1 , turn the real number χn+1 to
system control parameter and initial condition. its binary representation Xn+1 by Eq. (2), suppose that
Depending on the value of µ, the dynamics of the system L=45, thus Xn+1 is {b1,b2,b3,…b45}. By defining three
can change dramatically. The choice of µ in the equation variables whose binary representation is Xl=b1…b15,
above guarantees the system is in chaotic state and output Xm=b16…b30, Xh=b31…b45, respectively, the following
chaotic sequences {χn} have perfect randomness [25, 26]. equations are obtained.
A simple way for turning a real number χn to a discrete bits
symbol Xn is presented by Eq. (2). Change the decimal part X n +1′ = X l ⊕ X m ⊕ X h (4)
of the real into the binary sequences, and then extract Suppose that Xn+1′ is {p1…p15} after XOR operation,
some bits form it, so Xn is {bn1,bn2,bn3,…bnL}. We also we turn the Xn+1′ to its real representation χn+1′ by Eq. (5).
can turn the binary representation Xn to its corresponding χn+1′ is extended till tens to meet the condition:
real representation χn by the reverse operation of Eq. (2). χi+1′′∈(3.569945,4], so that the value can make the
xn = 0.bn1bn 2 ...bnL sequences into chaos. Then, judge whether the condition:
-1 -2 -L
(2) 3.569945< χi+1′′ ≤4 and c≥100 is valid or not, in it,
= 2 bn1 + 2 bn 2 + ... + 2 bnL variable c is the iterative times since last update of µ(2). If
(b) Keystream Generator valid, χi+1′′ is used to update the parameter µ(2) of the
second logistic map system. In the iterative process, χn+1′
The new generator system adopting two logistic maps is
proposed for the generation of pseudorandom binary sequences. is always to update the value of previous iteration χn.
This algorithm consists of two logistic maps: xn +1′ = 2 p1 + 2 p2 + ... + 2
-1 -2 -15
p15 (5)

© 2009 ACADEMY PUBLISHER

JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009 1093

B. Design of Encryption and Decryption Scheme

xn+1′′ = xn +1′ × 10 (6)
In this section, a chaos-based image encryption system,
= xi +1′′ (3.569945 < xi +1′′ ≤ 4 I c ≥ 100) (7) in the framework of stream cipher architecture, is
(2)
µ
proposed. The chaos-based image encryption scheme is
The second logistic map does the same operations shown in Fig. 2. An image is firstly converted to a binary
except for Eq. (6) and Eq. (7). Evolve successive states data stream. By masking these data with a random
from the second logistic map by yn+1=µ (2)yn(1-yn), and keystream generated by the chaos-based PRKG
obtain the real number yn+1 , turn the real number yn+1 to explained before, the corresponding encrypted image is
its binary representation Yn+1 by Eq. (2), then we can get formed. The details of encryption and decryption scheme
the value Yn+1′ by XOR operation expressed in Eq. (4), are to be discussed in the following Fig. 3. As
Yn+1′ is the output binary sequences zi, meanwhile, turn demonstrated in our simulation, this approach is
the Yn+1′ to its real representation yn+1′, and yn+1′ is to light-weighted and performed well both in security and
update the value of previous iteration yn. speed.
Briefly, the algorithm can be expressed as follows: As explained in Section 2.1, the PRKG is governed by
(1)
xi +1 = µ xi (1 − xi ) a couple of logistic maps, which is depended on the
1. values of (µ(1),χ0, µ(2),y0). These values are secreted, and
(2)
yi +1 = µ yi (1 − yi ), i = 0, 1, 2...
be used as the cipher key. Through iterations, the first
2. xi +1 → X i +1 , yi +1 → Yi +1 logistic map generates a hash value, which is highly
dependent on the input (i.e. the user key), is obtained and
X i +1′ = X l ⊕ X m ⊕ X h used to determine the system parameters of the second
3. logistic map. Then, the hash value will be enlarged till
Yi +1′ = Yl ⊕ Ym ⊕ Yh ten times. If the new obtained hash value is in the finite
area 3.569945<µ≤4, meanwhile, c≥100, the system
4. X n+1′ → xn+1′ , Yn+1′ → yn+1′ parameter of the second logistic map will be updated by
it. The second logistic map also generates a hash value,
5. xi +1′′ = xi +1′ × 10 which is highly dependent on the input (i.e. the user key)
and the first logistic map, is obtained and used to
= xi +1′′ (3.569945 < xi +1′′ ≤ 4 I c ≥ 100)
(2)
µ masking the data stream of the plain image. Because in
the plain image, the gray-scale values is in the range [0,
6. xi = xi +1′ , yi = yi +1′ 255], we implement the operation by Eq. (8).
7. zi = Yn +1′ zi′ = zi mod 256 (8)

Figure 2. Chaos-based image encryption scheme.

Figure 3. Chaotic encryption algorithm based on coupled chaotic maps.

© 2009 ACADEMY PUBLISHER

1094 JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009

shown in Fig. 4 (a) and Fig. 4 (c), where each image is in

III. SECURITY ANALYSIS AND TEST RESULTS 24-bit color with 256x256 pixels. By comparing the
original and the encrypted images in Fig. 4, there is no
In this section, the performance of the proposed image visual information observed in the encrypted image, and
encryption scheme is analyzed in detail. We discuss the the encrypted images are visual indistinguishable even
security analysis of the proposed image encryption
with a big difference in the color tone found in the
scheme including some important ones like statistical original images.
sensitivity, key sensitivity analysis, key space analysis In order to further demonstrate the effectiveness of our
etc. to prove the proposed cryptosystem is secure against
scheme, some more sophisticated tests suggested in [10,
the most common attacks. 11] have been carried out and the results are to be
A. Visual Testing explained in the followings.
A number of images are encrypted by the proposed
method, and visual test is performed. Two examples are

(a) (b)

(c) (d)
Figure 4. Frame (a) and (c) respectively, show the original images of Girl and Couple. Frame (b) and (d) respectively, show the encrypted images of
the plain images shown in frame (a) and (c) using the secret key’4000000000000, 1C28F5C28F5C,3F851EB851EB8, 333333333333 ’(in
hexadecimal).

and Fig. 4 (b), and their histograms of red, blue and

B. Statistical Analysis
green channels are shown in Fig. 5. It is clear that the
In order to resist the statistical attacks, the encrypted histogram of the encrypted image is nearly uniformly
images should possess certain random properties. We distributed, and significantly different from the
have performed statistical analysis by calculating the respective histograms of the original image. Hence the
histograms, the correlations of two adjacent pixels in the encrypted image does not provide any clue to employ
encrypted images and the correlation coefficient for any statistical attack on the proposed image encryption
several images and its corresponding encrypted images procedure, which makes statistical attacks difficult.
of an image database. A detail study has been undergone Experiments on various images have shown similar
and the results are summarized as followings. Different results. These properties tell that the proposed
images have been tested, and similar results are obtained. cryptosystem has high security against statistical attacks.
However, due to the page limit, only the results for the Compared with Fig. 1 in Ref.13, our results are more
Girl. (Fig. 4(a)) are used for illustration. persuasive. In the plain image, some gray-scale values in
(a) Histogram Analysis the range [0, 255] are not existed, but every gray-scale
In the experiments, the original image and its value in the range [0, 255] are existed and uniformly
corresponding encrypted image are shown in Fig. 4 (a) distributed in the encrypted image. In the Fig. 1 in Ref.13,

© 2009 ACADEMY PUBLISHER

JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009 1095

some gray-scale values are still not existed in the the same property is obtained, which also shows that the
encrypted image although the existed gray-scale values proposed chaos-based PRKG can generate chaotic
are uniformly distributed. Different images have been sequences with good pseudo-random properties.
tested by the proposed image encryption procedure, and

(a) (d)

(b) (e)

(c) (f)

Figure 5. Histogram analysis: Frames (a), (b) and (c) respectively, show the histograms of red, green and blue channels of the plain image shown in
Fig. 4 (a). Frames (d), (e) and (f) respectively, show the histograms of red, green and blue channels of the encrypted image shown in Fig. 4 (b).

© 2009 ACADEMY PUBLISHER

1096 JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009

Figure 6. Correlation of two adjacent pixels: Frames (a) and (b) respectively, show the distribution of two horizontally adjacent pixels in the plain and
encrypted images shown in Fig. 4 (a) and Fig. 4 (b). Frames (c) and (d) respectively, show the distribution of two vertically adjacent pixels in the
plain and encrypted images shown in Fig. 4 (a) and Fig. 4 (b).

(b) Correlation Coefficient Analysis Where x and y are the gray-scale values of two
adjacent pixels in the image and N is total number of
In order to test the correlation between two vertically
pixels selected from the image for the calculation. In the
adjacent pixels, and two horizontally adjacent pixels in
Table 1, we have given the correlation coefficients for
the several images and their encrypted images, we also
the original and encrypted images shown in Fig. 4 (a)
take Girl (Fig. 4 (a)) for example. In Fig. 6, we have
and Fig. 4 (b) respectively. In the Fig. 6 and Table 1, it is
shown the distribution of two adjacent pixels in the
clear that the two adjacent pixels in the original image
original and encrypted images shown in Fig. 4 (a) and
are highly correlated, but there is negligible correlation
Fig. 4 (b). Particularly, in Frames (a) and (b), we have
between the two adjacent pixels in the encrypted image.
depicted the distributions of two horizontally adjacent
Additionally, we have also done extensive study of the
pixels in the original and encrypted images respectively.
correlation between image and its corresponding
Similarly, in Frames (c) and (d) respectively, the
encrypted image by using the proposed encryption
distributions of two vertically adjacent pixels in the
algorithm. We have used the USC-SIPI image database
original and encrypted images have been depicted.
which is referred in Ref.13 (freely available at
Compared with Fig. 2 in Ref.13, distribution of two
http://sipi.usc.edu/database/).
adjacent pixels are more uniform in our results, but in the
The secret key ’4000000000000, 1C28F5C28F5C,
Fig. 2 in Ref.13, distribution displays concentrated trend.
3F851EB851EB8, 333333333333 ’(in hexadecimal) has
Moreover, we have also calculated the correlation
been used for encryption process. Results are shown in
between two vertically as well as horizontally adjacent
the Table 2. The correlation coefficient is very small
pixels in the original encrypted images, using the
which implies that no correlation exists between original
following formula referred in Ref.13.
N N N
and its corresponding encrypted images. Compared with
N ∑ (x j × y j ) − ∑ x j × ∑ y j Table3 in Ref.13, the correlation coefficient between
j =1 j =1 j =1 (8) original and its corresponding encrypted images are
Cr =
N N N N generally smaller in our results.
( N ∑ x j 2 − (∑ x j ) 2 ) × ( N ∑ y j 2 − (∑ y j ) 2 )
j =1 j =1 j =1 j =1

Table 1. Correlation coefficients for the two adjacent pixels in the original and encrypted images shown in Fig. 4.
Original image (Fig.4(a)) Encrypted image (Fig.4(b))
Horizontal 0.9635 -0.0030
Vertical 0.9711 -0.0028

© 2009 ACADEMY PUBLISHER

JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009 1097

Table 2. Correlation coefficients between the image and corresponding encrypted image for a number of images of the USC-SIPI image database. The
encryption has been done using the secret key ’4000000000000, 1C28F5C28F5C, 3F851EB851EB8, 333333333333 ’.
File name File description Size Type Correlation coefficient Correlation coefficient
between the image and for the two adjacent
corresponding pixels in the encrypted
encrypted image image
4.1.01 Girl 256x256 Color -0.000492 -0.002887
4.1.02 Couple 256x256 Color -0.001949 -0.002184
4.1.03 Girl 256x256 Color 0.001619 -0.000796
4.1.04 Girl 256x256 Color 0.001906 -0.003210
4.1.05 House 256x256 Color 0.001563 -0.001986
4.1.06 Tree 256x256 Color -0.001921 -0.003900
4.1.07 Jelly beans 256x256 Color 0.000114 -0.002231
4.1.08 Jelly beans 256x256 Color 0.002656 -0.002368
4.2.01 Splash 512x512 Color 0.001096 -0.001326
4.2.02 Girl(Tiffany) 512x512 Color -0.001220 -0.002239
4.2.03 Baboon 512x512 Color 0.000744 -0.000542
4.2.04 Girl(lenna) 512x512 Color 0.001857 -0.000983
4.2.05 Airplane(F-16) 512x512 Color 0.001911 -0.001005
4.2.06 Sailboat on lake 512x512 Color 0.002089 0.000037
4.2.07 Peppers 512x512 Color 0.002248 -0.001883
House House 512x512 Color 0.000607 -0.000728
5.1.09 Moon surface 256x256 Gray 0.000265 -0.001291
5.1.10 Aerial 256x256 Gray 0.004396 -0.002667
5.1.11 Airplane 256x256 Gray 0.007487 -0.000216
5.1.12 Clock 256x256 Gray 0.014194 -0.002033
5.1.13 Resolution chart 256x256 Gray 0.006588 0.001655
5.1.14 Chemical plant 256x256 Gray 0.000024 -0.003028
5.2.08 Couple 512x512 Gray 0.003597 -0.000508
5.2.09 Aerial 512x512 Gray 0.003001 -0.002298
5.2.10 Stream and bridge 512x512 Gray 0.002980 -0.000530
7.1.01 Truck 512x512 Gray 0.001955 -0.001081
7.1.02 Airplane 512x512 Gray 0.003604 -0.001460
7.1.03 Tank 512x512 Gray 0.002931 0.000002
7.1.04 Car and APCs 512x512 Gray 0.001468 -0.001481
7.1.05 Truck and APCs 512x512 Gray 0.004383 -0.001513
7.1.06 Truck and APCs 512x512 Gray 0.003054 -0.002319
7.1.07 Tank 512x512 Gray -0.000167 -0.002164
7.1.08 APC 512x512 Gray 0.000198 -0.001414
7.1.09 Tank 512x512 Gray 0.003317 -0.002334
7.1.10 Car and APCs 512x512 Gray 0.000862 -0.003146
boat.512 Fishing Boat 512x512 Gray -0.000342 0.000087
elaine.512 Girl(Elaine) 512x512 Gray 0.002869 -0.001650
gray21.512 21 level step wedge 512x512 Gray 0.001652 -0.000524
numbers.512 256 level test 512x512 Gray 0.001638 -0.002177
ruler.512 Pixel ruler 512x512 Gray 0.000079 -0.000790
5.3.01 Man 1021x1024 Gray 0.000413 -0.000102
5.3.02 Airport 1021x1024 Gray 0.001564 -0.000221
7.2.01 Airplane 1021x1024 Gray 0.001389 0.000430
testpat.1k General test pattern 1021x1024 Gray 0.001466 -0.000247

C. Sensitivity Analysis corresponding encrypted image is referred as encrypted

image A (Fig. 7 (b)).
An encryption scheme has also to be key-sensitive,
(2) Then, the least significant bit of the key is changed,
meaning that a tiny change in the key will cause a
so that the original key becomes key2
significant change in the output. In our tests, we use the
‘µ(1)=3F9F6C38FB144, µ(2) =3F9F6C38FB144’, which is
fixed initial value ‘χ0=1C28F5C28F5C,
used to encrypt the same image, and its corresponding
y0=333333333333’, changing the system parameter ‘µ(1),
encrypted image is referred as encrypted image B (Fig. 7
µ(2)’with a single bit. We know that the system parameter
(c)).
can be any value in the finite area 3.569945<µ≤4, thus
(3) Again, the same image is encrypted by the key3
we can provide µ(1) and µ(2) with the same value. The key
‘µ(1)=3FAF6C38FB143, µ(2) =3FAF6C38FB143’, and its
sensitivity test is performed in detail according to the
corresponding encrypted image is referred as encrypted
following steps:
image C (Fig. 7 (d)).
(1) First, a 256x256 image (Fig. 7 (a)) is encrypted by
(4) Finally, the above three encrypted images A, B and
using the test key1 ‘µ(1)=3F9F6C38FB143, µ(2)
C, encrypted by the three slightly different keys, are
=3F9F6C38FB143’ (in hexadecimal), and its
compared.

© 2009 ACADEMY PUBLISHER

1098 JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009

(a) (b)

(c) (d)
Figure 7. Key sensitivity test 1: Frame (a): plain image; Frame (b): encrypted image A by key1 ‘µ(1)=3F9F6C38FB143, µ(2) =3F9F6C38FB143’;
Frame (c): encrypted image B by key2 ‘µ(1)=3F9F6C38FB144, µ(2) =3F9F6C38FB144’; Frame (d): encrypted image C by key3
‘µ(1)=3FAF6C38FB143, µ(2) =3FAF6C38FB143’.

Table 3. Correlation coefficients between the plain image and the three different encrypted images obtained by using slightly different secret keys.
Image 1 Image 2 Correlation
coefficient
Plain image (Fig.7 (a)) Encrypted image A（Fig.7 (b)） -0.005490
Plain image(Fig.7 (a)) Encrypted image B（Fig.7 (c)） 0.003207
Plain image(Fig.7 (a)) Encrypted image C（Fig.7 (d)） 0.001584
Table 4. Correlation coefficients between the corresponding pixels of the three different encrypted images obtained by using slightly different secret
key of an image.
Image 1 Image 2 Correlation Difference pixels
coefficient (mean NPCR)
Encrypted image A（Fig.7 (b)） Encrypted image B（Fig.7 (c)） 0.000246 99.611%
Encrypted image B（Fig.7 (c)） Encrypted image C（Fig.7 (d)） 0.001116 99.623%
Encrypted image C（Fig.7 (d)） Encrypted image A（Fig.7 (b)） -0.000474 99.605%

For comparison, we use the same ways which is keys. In Table 4, it can be observed that the values are
referred in Ref.13, using the same formula as given in very close to the expected value of pixel difference on
Eq.8 except that in this case x and y are the values of two randomly generated images (99.609375%). We also
corresponding pixels in the two encrypted images to be obtained NPCR for a large number of images by using
compared. In table 3 and 4, as can be seen, the our encryption scheme, and found the same results that
correlation between a plain image and the corresponding the encryption scheme is very sensitive with respect to
encrypted images is negligible around zero, which shows small changes in the key.
that the plain image is nearly independent from the Moreover, if we use a trivially modified key to
encrypted images. This is consistent with the perfect decrypt the ciphered image, then the decryption should
security defined by Shannon [27]. Similarly, the not succeed. Figure 8 has verified this, where the image
correlation between different encrypted images is also encrypted by the key1 was not be correctly decrypted by
negligible around zero, which shows that the encrypted using the key2 and key3. Here, there is only one bit
images are independent from each other. We have difference between the three keys explained before.
measured the number of pixel change rate (NPCR) of Those results clearly show high key-sensitivity of the
two cipher images with only one bit difference in the proposed chaotic encryption algorithm.

© 2009 ACADEMY PUBLISHER

JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009 1099

(a) (b) (c)

(d) (e)
Figure 8. Key sensitivity test 2: Frame (a) is the plain image, with its encrypted image Frame (b) by key1 ‘µ(1)=3F9F6C38FB143, µ(2)
=3F9F6C38FB143’. Frame (c), (d), and (e) respectively, show the decrypted images by the key1 ‘µ(1)=3F9F6C38FB143, µ(2) =3F9F6C38FB143’,
key2 ‘µ(1)=3F9F6C38FB144, µ(2) =3F9F6C38FB144’, and key3 ‘µ(1)=3FAF6C38FB143, µ(2) =3FAF6C38FB143’ to decrypt encrypted image Frame
(b).

statistical analysis on the proposed encryption scheme is

D. Key Space Analysis
given. From the experimental results, it is concluded that
Key space size is the total number of different keys it outperforms existing schemes, both in terms of speed
that can be used in the encryption. Cryptosystem is and security. Having a high throughput, the proposed
completely sensitive to all secret keys. A good system is ready to be applied in fast real time encryption
encryption algorithm should not only be sensitive to the applications.
cipher key, but also the key space should be large enough
to make brute-force attack infeasible. If the precision is ACKNOWLEDGEMENTS
10-15, the key space size for initial conditions and control
parameters is over than 2196. Apparently, the key space is The work was supported by the National Basic
sufficient for reliable practical use. Research Program of China (Grant No. 2006CB303104)
and the National Natural Science Foundation of China
E. Time Analysis (Grant No. 40871200).
All the security analysis has been done on MATLAB
7.0 by AMD Athlon(tm) 64 X2 Dual Core processor REFERENCES
3800+ 2.0GHz personal computer. We also do the time
[1] B. Hao, Starting with parabolas: an introduction to
analysis on it, but not ideal. The keysteam generation chaotic dynamics, Shanghai Scientific and Technological
system consumes longer time. Considering that the two Education Publishing House, Shanghai, China, 1993.
chaotic logistic maps are suitable for parallel computing [2] R. Brown, L.O. Chua, “Clarifying chaos: examples and
in hardware, 64-bit precision representation with fixed counterexamples,” Int. J. Bifurcat Chaos, vol. 6, pp.
point arithmetic is assumed, and a fast throughput and 219–242, 1996.
facilitate VLSI architecture is implemented on ALTERA [3] J. Fridrich, “Symmetric ciphirs based on two-dimensional
QUATUS II 5.0. The keystream output speed is up to chaotic maps,” Int. J. Bifurcat Chaos, vol. 8, pp. 1259–
571.429 Mbps [28], which is strongly suitable for the use 1284, 1998.
[4] S. Li, Q. Li, W. Li, X. Mou, Y. Cai, “Statistical properties
of most of real-time video and audio applications.
of digital piecewise linear chaotic maps and their roles in
cryptography and pseudo-random coding,” IMA Int Conf
IV. CONCLUSIONS Crypt & Coding, vol. 2260, pp. 205–221, 2001.
In this paper, an image encryption scheme based on [5] B. Schneier, Applied Cryptography–Protocols,
Algorithms, and Source Code in C, 2nd edition, John
coupled chaotic logistic maps is proposed. The system is Wiley & Sons, Inc., New York, 1996.
in a stream-cipher architecture, where the PRKG is [6] J. Daemen, B. Sand, V. Rijmen, The Design of Rijndael:
formed by two chaotic maps, serving the purpose of AES–The Advanced Encryption Standard,
stream generation and random mixing, respectively. It is Springer-Verlag, Berlin, 2002.
found that such a design can enhance the randomness, [7] S. Li, Q. Li, X. Mou, Y. Cai, “Pseudo-random bit
even under finite precision implementation. A detailed generator based on couple chaotic systems and its
applications in stream-cipher cryptography,” LNCS 2247,

© 2009 ACADEMY PUBLISHER

 1100 JOURNAL OF COMPUTERS, VOL. 4, NO. 11, NOVEMBER 2009

pp. 316–329, 2001. Management, vol.2, Gyeongju, KOREA, 2008.

[8] S. Li, X. Zheng, “Cryptanalysis of a chaotic image
encryption method,” in: Proceedings of the IEEE Shubo Liu received the B.E. degree in
International Symposium on circuits and systems, electronic technology and the M.S. degree
Scottsdale, AZ, USA, 2002 in computer application from Wuhan
[9] K.W. Wong, S.W. Ho, C.K. Yung, “A chaotic University (WHU), Hubei, China, in 1993
cryptography scheme for generating short ciphertext,” and 1999, respectively, and he is currently
Phys. Lett. A, vol. 310, pp. 67–73, 2003. pursuing the Ph.D. degree in
[10] G. Chen, Y. Mao, C.K. Chui, “A symmetric image communication and information system at
encryption based on 3D chaotic maps,” Chaos Solitons WHU.
Fractals, vol. 21, pp. 749–761, 2004. Since 1993, he has been on the faculty
[11] Y.B. Mao, G. Chen, S.G. Lian, “A novel fast image in computer school at WHU, and he is
encryption scheme based on the 3D chaotic baker map,” currently an Associate Professor in the Department of
Int. J. Bifurcat Chaos, vol. 14, pp. 3613–3624, 2004. Computer Application. His research interests include
[12] S.G. Lian, J. Sun, Z. Wang, “A block cipher baser on a multimedia information security, and multimedia network
suitable use of chaotic standard map,” Chaos Solitons communication technology.
Fractals, vol. 26, pp. 117–129, 2005. Mr. Liu was awarded two scientific and technological
[13] N.K. Pareek, V. Patidar, K.K. Sud, “Image encryption progress prizes and one invent patent applying.
using chaotic logistic map,” Image and Vision Computing,
vol. 24, pp. 926–934, 2006. Jing Sun received the B.E. degree in
[14] T. Xiang, X. Liao, G. Tang, Y. Chen, K.W. Wong, “A information security form Hubei Police
novel block cryptosystem based on iterating a chaotic University, China, in 2007, and the M.S.
map,” Phys. Lett. A, vol. 349, pp. 109–115, 2006. degree in computer application form
[15] H.S. Kwok, W.K.S. Tang, “A fast image encryption Wuhan University (WHU), China. She is
system based on chaotic maps with finite precision currently pursuing the Ph.D. degree in
representation,” Chaos Solitons Fractals, vol. 32, pp. communication and information system at
1518–1529, 2007. WHU.
[16] T. Xiang, K.W. Wong, X. Liao, “A novel symmetrical She is a researcher in the Multimedia
cryptosystem based on discretized two-dimensional Information Security of the State Key
chaotic map,” Phys. Lett. A, vol. 364, pp. 252–258, 2007. Laboratory of Information Engineering in Surveying, Mapping
[17] P. Li, Z. Li, W.A. Halang, G. Chen, “A stream cipher and Remote Sensing at WHU. Her research interests include
based on a spatiotemporal chaotic system,” Chaos multimedia information security, information theory, and
Solitons Fractals, vol. 32, pp. 1867–1876, 2007. communication networks.
[18] J. Wei, X. Liao, K.W. Wong, T. Zhou, “Cryptoanalysis of Miss. Sun received the Master First-Level Scholarship (Best
a cryptosystem using multiple one-dimensional chaotic Student Award) form WHU, in 2008.
maps,” Commun. Nonlinear Sci. Numer. Simul, vol. 12,
pp. 814–822, 2007. Zhengquan Xu received the B.E. degree in
[19] K.W. Wong, B.S.H. Kwok, W.S. Law, “A fast image radio technology and information system
encryption scheme based on chaotic standard map,” Phys. and M.S. degree in communication and
Lett. A, vol. 372, pp. 2645–2652, 2008. electronic system form Tsinghua University,
[20] S. Behnia, A. Akhshani, H. Mahmodi, A. Akhavan. “A China, in 1985 and 1988, respectively, and
novel algorithm for image encryption based on mixture of the Ph.D degree in biomedicine engineering
chaotic maps,” Chaos Solitons Fractals, vol. 35, pp. form Hong Kong Polytechnic University, in
408–419, 2008. 1998.
[21] N.K. Pareek, Vinod Patidar, K.K. Sud, “Discrete chaotic From 1988 to 1999, he was on the
cryptography using external key,” Phys. Lett. A, vol. 309, faculty in the department of Electronics and Information at the
pp. 75–82, 2003. Huazhong University of Science and Technology. In 1995, he
[22] N.K. Pareek, Vinod Patidar, K.K. Sud, “Cryptography was selected to the Hong Kong Polytechnic University (HKPU)
using multiple one-dimensional chaotic maps,” Commun. by the State Education Ministry. During the period, he was
Nonlinear Sci. Numer. Simul, vol.10, pp. 715–723, 2005. selected to Stanford University as a short-term visiting scholar
[23] M.A. Jafarizadeh, S. Behnia, S. Khorram, H. Nagshara, by the “Areas of Excellence” plan of the HKPU. Since 1999, he
“Hierarchy of chaotic maps with an invariant measure,” J has been on the faculty in Wuhan University (WHU). He is
Stat Phys, vol. 104, pp. 1013–1028, 2001. currently the director of multimedia communication
[24] M.A. Jafarizadeh, S. Behnia, “Hierarchy of chaotic maps engineering center of state key laboratory of information
with an invariant and their coupling,” Physica D, vol. 159, engineering in surveying, mapping and remote sensing at WHU.
pp.1–21, 2001. He is the expert of government affairs information construction
[25] M.S. Baptista, “Cryptography with chaos,” Phys. Lett. A, engineering of Wuhan City, and the expert of government
vol. 240, pp. 50–54, 1998. affairs information construction engineering of Hubei Province.
[26] T. Geisel, V. Fairen, “Statistical Properties of Chaos in His research interests include space and multimedia
Logistic Map,” Phys. Lett. A, vol. 105, pp. 263-266, 1984. information processing, multimedia information security, and
[27] C. Shannon, “Communication theory of secrecy systems,” multimedia network communication technology.
Bell Syst Tech J, vol. 28, pp. 656–715, 1949. Mr. Xu has presided or participated in a variety of more than
[28] S.B. Liu, J. Sun, Z.Q. Xu, Z.H. Cai, “An improved 20 research projects and was awarded five scientific and
chaos-based stream cipher algorithm and its VLSI technological progress prizes: three Ministerial-level and two
implementation”, 4th International Conference on Municipal-level awards respectively.
Networked Computing and Advanced Information

© 2009 ACADEMY PUBLISHER

View publication stats