chaotic dynamical systems based designing of nonlinear block
TRANSCRIPT
Chaotic Dynamical Systems based Designing of Nonlinear Block
Cipher Component and Their Applications to Multimedia Security
Adnan Javeed
Department of Mathematics
Quaid-i-Azam University
Islamabad, Pakistan
2020
Chaotic Dynamical Systems based Designing of Nonlinear Block
Cipher Component and Their Applications to Multimedia Security
Adnan Javeed
Supervised
By
Prof. Dr. Tariq Shah
Department of Mathematics
Quaid-i-Azam University
Islamabad, Pakistan
2020
Chaotic Dynamical Systems based Designing of
Nonlinear Block Cipher Component and Their
Applications to Multimedia Security
A Thesis Submitted to the Department of Mathematics,
Quaid-i-Azam University, Islamabad, in the partial fulfillment of
the requirement for the degree of
Doctorate of Philosophy
in
Mathematics
By
Adnan Javeed
Department of Mathematics
Quaid-i-Azam University
Islamabad, Pakistan
2020
Authorโs Declaration
I Adnan Javeed hereby state that my PhD thesis titled Chaotic Dynamical
Systems based Designing of Nonlinear Block Cipher Component and
Their Applications to Multimedia Security is my own work and has not
been submitted previously by me for taking any degree from the Quaid-I-Azam
University Islamabad, Pakistan or anywhere else in the country/world.
At any time if my statement is found to be incorrect even after my graduation,
the university has the right to withdraw my PhD degree.
Name of Student: Adnan Javeed
13-09-2020
Acknowledgment
All praise for Almighty Allah, the Creator and the Merciful Lord, who guides me in
darkness, helps me in difficulties and enables me to reach the ultimate stage with
courage.
I express the deepest gratitude to my respected supervisor Prof. Dr. Tariq Shah for
his guidance, constant encouragement and suggestions throughout my research work.
In short, his tireless work and unique way of research cannot be expressed in words. I
am also thankful to the respected Chairman, Department of Mathematics, Prof. Dr.
Sohail Nadeem for his support and guidance.
I wish to express my heartiest thanks and gratitude to my parents Mr. and Mrs. Malik
Javed Akhtar, the ones who can never ever be thanked enough for the overwhelming
love, kindness and care they bestow upon me. I am also thankful to each member of
my family. Special thanks to my wife and kids.
I would like to express my gratitude to all the respected teachers. They are all those
people who made me what I am today, they polished me at different stages of my life
and taught me whatever I am today.
I enact my heartiest and deepest thanks to all my lab fellows and especially the
research fellows Dr Attaullah and Dr Jameel who were there during my Ph. D session.
I gratefully acknowledge my seniors, my friends and Ph. D fellows for their brilliant
ideas and important contribution in refining my research work. Their professional
guidance has nourished my skills and I will always remain thankful to them.
I am also grateful to the administrative staff of mathematics department for their
support at every time.
In the end thanks to all my research fellows and to those people who directly and
indirectly helped me during my research work.
Adnan Javeed
13-09-2020
Preface
The absolute achievements of information sciences in last few decades are extensive
deployment of soft and small computing devices in general public along with the speedy
communication channel. An easy approach to valuable digital data had to face some security
apprehensions. Frequent transmission and communication of information bears problems like
copyright protection, false ownership claims and alteration in valued information, integrity,
confidentiality, non-repudiation, access control and authenticity. All of these and many more
of this type of security issues are the matter of concern for researchers as well as for officials.
The security of data is preserved in such vulnerable situations by making use cryptography.
Cryptography works generally on mechanisms of converting meaningful information into non
readable form and vice versa. The two main divisions of cryptography are symmetric and
asymmetric key cryptography. These two branches are bifurcated on the basis of keys. Same
key is used for encryption and decryption in symmetric key cryptography whereas different
set of keys are used in both these procedures in asymmetric cryptosystems. Stream ciphers
and block ciphers are the two broad categories of symmetric key cryptography. In block
cryptograms, the procedure of enciphering is done for the blocks of data with different sizes.
The only nonlinear and complex part of block cryptosystem capable of generating hurdles for
cryptanalysts is the substitution box (S-box).
After the development of advanced encryption standard, the need of new encryption standard
is diminished because of its robustness and strength against cryptanalyses. However, its
security can be enhanced by using chaos based substitution boxes (S-boxes) instead of
algebraic S-boxes. S-box is the nonlinear component of block cipher responsible for creating
confusion in the system. It can have different dimensions depending upon the need of
algorithm. It is produced in the form of square matrix from a mathematical structure.
Nonlinear mathematical systems are the suitable candidates for the generation of S-boxes.
In literature, large number of articles are available expressing research work of scientists
related to cryptography and chaos. This list starts from the algebraic structures, one
dimensional chaotic systems and moves on towards complex chaotic circuits. One
dimensional systems are easy to manipulate because of their simplicity and plainness. But
they are more vulnerable to cyber threats like brute force and linear attacks due to the low
key space, small chaotic range and involvement of fewer number of variables. Issues related
to algebraic structures and circuits are computational complexity and slow processing speed.
These drawbacks motivated many cryptographers to use nonlinear chaotic systems in the
place of simple schemes. Differential equations based chaotic dynamical systems hamper all
such deficiencies and threats.
Differential equations is a nurtured branch of mathematics which has the ability to model any
dynamical system changing its position or characteristics etc. Chaotic dynamical systems can
be divided into two classes namely, discrete and continuous. Dynamical systems based on
differential equations fall in the category of continuous chaotic dynamical system. These can
further be classified based on the order of the differential equations, nature of the derivatives
like ordinary/partial, or by the system of differential equations. Chaos observance in all cases
is possible and there is no proper rule that can be assigned to a physical system for the
generation of chaos. Its occurrence is totally unpredictable.
This thesis primarily focuses on the generation of S-boxes from chaotic dynamical system
and second aim is to design encryption/watermarking techniques by using chaotic system of
differential equation. At the first stage of thesis, chaos and cryptography are mainly
discussed. Some basic terminology of cryptography and block ciphers are elaborated in
detail. Moreover, important properties of S-box on the perspective of mathematics are also
elucidated. Secondly, first order system of coupled differential equations generating chaotic
and complex bifurcation pattern are identified and solved numerically. The sensitivity of such
system on initial conditions and parameters was examined keenly to authenticate its chaotic
nature. The solution space of the coupled differential equations is then utilized for the
construction of S-boxes. One S-box generated from this system is evaluated using standard
algebraic and statistical analyses to prove its efficacy for further utilization in image
encryption and water marking.
The simplest chaotic dynamical system is double pendulum. Chapter 3 gave an idea of
designing nonlinear component of block cipher using double pendulum and the permutation
group. Symmetric group is used for permutation at the final stage. The result analyses are
promising for its further use in multimedia security.
An application of Rabinovich-Fabrikant (RF) system in the form of image encryption and
watermarking is given in chapter four. Chaotic solution of RF system with different initial
conditions and parameters is achieved using Runge-Kutta method of order four in MATLAB.
Three dissimilar sequences from the solution space are extracted for their further utilization
in colour image encryption. Each channel of red, green and blue are separately operated to
scramble the original positioning of pixels to yield an encrypted meaningless image.
Moreover, the same sequences are implemented for watermarking which was done along
each channel. The simulation results of all major standard analyses were observed carefully
and found clear in determining the robustness of the proposed schemes of encryption and
watermarking.
In chapter five, a second order differential equation also known as Duffing equation is
debated. The chaotic behaviour of this equation is seen in the form of oscillator. The
interesting attribute of this oscillator is the production of different bifurcation pattern for
slight variation in parameters/conditions causing chaos. Its density distribution function is
complex, rich and dense. It means one can generate a large number of S-boxes from this
single system. Lightweight but yet practical and applicable image encryption based on
chaotic differential equation is presented in this chapter. All the standard analyses were found
promising in analysing the suggested scheme of encryption. This thesis has been ended with
chapter 6 which includes the conclusions and the future directions.
1
Contents
Chapter 1 .................................................................................................................................. 5
Chaos and Cryptography: An Introduction .......................................................................... 5
1.1 Introduction ............................................................................................................... 5
1.2 Research Objectives .................................................................................................. 8
1.3 Thesis Layout ............................................................................................................. 9
1.4 Cryptography and Chaos ....................................................................................... 11
1.4.1 Chaotic System ................................................................................................. 11
1.4.2 Chaotic Dynamical Systems ............................................................................ 12
1.4.3 Causes of a Chaotic System ............................................................................. 13
1.4.4 Characteristics of a Chaotic System ............................................................... 13
1.4.5 One Dimensional Discrete Chaotic System.................................................... 14
1.5 Differential Equations Based Chaotic Dynamical System........................................ 17
1.5.1 Rossler Attractor .............................................................................................. 17
1.5.2 Lorenz System .................................................................................................. 17
1.5.3 Rabinovich-Fabrikant System of Equations ................................................. 17
1.5.4 Double Pendulum ............................................................................................. 18
1.5.5 Second order Differential Equation (Duffing Oscillator)............................. 18
1.6 Boolean Algebra ...................................................................................................... 19
1.6.1 Boolean Function ............................................................................................. 20
1.6.2 Hamming Weight ............................................................................................. 20
1.6.3 Hamming Distance ........................................................................................... 20
1.6.4 Correlation........................................................................................................ 21
1.6.5 Algebraic Normal Form (ANF) ...................................................................... 21
1.6.6 Walsh Hadmard Transform ........................................................................... 22
2
1.7 Cryptographic Properties of Substitution box .......................................................... 22
1.7.1 S-Box ................................................................................................................. 22
1.7.2 Nonlinearity ...................................................................................................... 22
1.7.2 Bit Independence Criterion ............................................................................. 23
1.7.3 Strict Avalanche Criterion (SAC) .................................................................. 23
1.7.4 Linear Approximation Probability ................................................................ 24
1.7.5 Differential Approximation Probability ........................................................ 24
1.7.6 Majority Logic Criterion ................................................................................. 24
1.8 Randomness Test (NIST SP-800 22) ........................................................................... 25
Chapter 2 ................................................................................................................................ 26
First Order Coupled Differential Equations based Dynamical System for the
Construction of S- boxes ........................................................................................................ 26
2.1 Background .............................................................................................................. 26
2.2 Coupled Differential Equations Based Construction of S-box ........................... 28
2.3 Algebraic and Statistical Result Analyses of S-box .............................................. 32
2.3.1 Nonlinearity ...................................................................................................... 32
2.3.2 Bit Independence Criterion ............................................................................. 32
2.3.3 Strict Avalanche Criterion .............................................................................. 33
2.3.4 Linear Approximation Probability ................................................................ 33
2.3.5 Differential Approximation Probability ........................................................ 34
2.3.6 Majority Logic Criteria ................................................................................... 34
Chapter 3 ................................................................................................................................ 38
Construction of Non-Linear Component of Block Cipher by means of Chaotic Dynamical
System and Symmetric Group .............................................................................................. 38
3.1. Contextual Review................................................................................................... 38
3.1. Double Pendulum .................................................................................................... 39
3.2. Construction of S-Box ............................................................................................. 43
3.3. Analysis of S-box ..................................................................................................... 46
3.4 Majority Logic Criterion ............................................................................................. 49
3
Chapter 4 ................................................................................................................................ 52
Applications of Nonlinear Coupled Differential Equations in Multimedia Security ...... 52
4.1 Introduction ............................................................................................................. 52
4.2 RabinovichโFabrikant Equations .......................................................................... 54
4.2.1. The Literature Review ..................................................................................... 55
4.3 Proposed Image Encryption Scheme ..................................................................... 56
4.4 Security Standards to gauge the strength ............................................................. 58
4.4.1. Key Space .......................................................................................................... 58
4.4.2. Key Sensitivity Analysis .................................................................................. 59
4.4.3. Complexity Analysis ........................................................................................ 60
4.4.4. Information Entropy ....................................................................................... 61
4.4.5. Correlation Analysis ........................................................................................ 61
4.4.6. Histogram Analysis .......................................................................................... 63
4.4.7. Sensitivity Analysis .......................................................................................... 64
4.4.8 Randomness Test (NIST SP 800-22) for Cipher ........................................... 66
4.5 Watermarking Scheme ........................................................................................... 68
4.5.1 Embedding Process .......................................................................................... 68
4.5.2 Extraction Process ........................................................................................... 69
4.6 Simulation Results and Statistical Analysis .......................................................... 71
4.6.1 Mean Squared Error (MSE) ........................................................................... 71
4.6.2 Peak Signal to Noise Ratio (PSNR) ................................................................ 72
4.7 Robustness Test on Image Processing Operations ............................................... 73
4.7.1 Noise Attack ...................................................................................................... 73
4.7.2 Cropping Attack............................................................................................... 74
Chapter 5 ................................................................................................................................ 76
Lightweight Secure Image Encryption Scheme Based on Second Order Chaotic
Differential Equation ............................................................................................................. 76
5.1 Introduction ............................................................................................................. 76
5.2 Chaotic Duffing Oscillator ..................................................................................... 79
4
5.3 Substitution Box Based on Chaotic Dynamical System ....................................... 81
5.4 Algebraic Strength of an S-box .............................................................................. 82
5.4.1 Nonlinearity ...................................................................................................... 83
5.4.2 Bit Independence Criterion ............................................................................. 83
5.4.3 Strict Avalanche Criterion .............................................................................. 84
5.4.4 Differential Approximation Probability ........................................................ 84
5.4.5 Linear Approximation Probability ................................................................ 85
5.5 Proposed Scheme for Image Encryption ............................................................... 85
5.6 Investigational Upshots and Simulation Analyses ............................................... 88
5.6.1 Key Space Analysis .......................................................................................... 88
5.6.2 Key Sensitivity Analysis .................................................................................. 89
5.6.3 Correlation Analysis ........................................................................................ 89
5.6.4 Histogram Analysis .......................................................................................... 90
5.6.5 Information Entropy ....................................................................................... 92
5.6.6 Sensitivity Analysis .......................................................................................... 92
5.7 Randomness Test (NIST SP 800-22) for cipher ......................................................... 94
Chapter 6 ................................................................................................................................ 97
Conclusion and Future Directions ........................................................................................ 97
6.1 Conclusion ................................................................................................................ 97
6.2 Future Work ............................................................................................................ 99
References ............................................................................................................................. 101
5
Chapter 1
Chaos and Cryptography: An Introduction
The objective of secure communication in todayโs world is well defined goal of every
communicating party. The primitive idea in attaining this objective is the generation of
nonlinear component of block cipher designed with the help of chaotic dynamical system is the
main concern of this thesis. In this chapter some brief discussion about chaos and cryptography
is given right after highlighting the objective and structure of the thesis. Secondly, substitution
box S-box and its cryptographic properties along with the basics of cryptography are given.
1.1 Introduction
A huge amount of digital content communicated through the insecure channel is the tip of the
iceberg. Breach of sensitive and treasured information at times become disastrous for the social
order. Digital data being transmitted is under the access of the channel responsible for
transmission. Insecure channel is indeed a matter of concern for a large number of populaces.
There are many problems related to the security of the digital data but the crux of the matter is
the insecure channel involved in the communications. The optimal solution to this problem is
to build reliable and secure channel.
An immediate substitute for the secure transmission of digital data is the utilization of
cryptographic algorithms. Cryptography being a vast domain offers many vibrant and robust
cryptosystems. It assures confidentiality, integrity and authentication of the data. The main goal
is to bypass the media between the receiver and sender by converting valuable information into
6
a bogus file. It involves recurring and logical based mathematical operations, which makes the
whole process to generate a vicious circle.
Cryptography is the field of science which deals with the techniques for data fortification,
copyright protection and safe communication via insecure channels. It is accomplished with
the subject knowledge of computer science and mathematics by designing algorithms that uses
reversible schemes to hide the information for secure transmission. The recovery of the original
information is only possible if an accurate set of keys are applied while cryptanalyzing the
cryptosystems. Based on keys, there are two main classification of cryptography i.e. symmetric
key cryptography and asymmetric key cryptography.
Same key is used in symmetric algorithms for encryption and decryption whereas different
keys are used for encryption and decryption in asymmetric key cryptography, also known as
public key cryptography. The input data in the form of blocks and streams further divide the
ciphers into block and stream ciphers respectively. Additionally, hash function being another
cipher works by squashing the input information.
Confusion and diffusion introduced in [1] are the two fundamental attributes of a robust
cryptosystem. Former is achieved by the ambiguous relationship of each single binary bit with
the key. While, later proposes that fifty percent of output bits must change whenever sole input
bit is complemented. It is preferred to design a cryptosystem whose strength can be enhanced
by slightly altering parameters which is accomplished by the generation of confusion and
diffusion in that system. Boolean functions are the one example satisfying the above criteria
hence their presence in such systems is mandatory.
After the invention of advanced encryption standard (AES) [2], the need of designing new
standard is minimized since invention itโs unbreakable. The only nonlinear component of block
cipher AES is S-box. Its construction is purely of mathematical background. This motivated
many people to utilize several mathematical systems to design cryptographically strong S-
7
boxes to enhance the security of a cryptosystem instead of designing new encryption standard.
The goal of attaining confusion in block cipher is due to the action of S-box on the cipher text.
Moreover, S-boxes are also utilized in designing steganography, watermarking and image
encryption see [3], [4] and [5].
Boolean function and Block cipherโs nonlinear component (S-box) are recognized as the
imperative parts of a modern cryptosystem. Former gives single output for a solo input while
later generates multiple output bits for a single input bit. Both are linked with each other by the
application of theory of functions.
S-box is a look up table constructed from a mathematical system. Initially, the plain-text is
divided in the form of blocks of information in bytes. This look up table is used to substitute
the original valued information by a meaningless entry from S-box. The chance of recognizing
original bits (bytes) is almost zero whenever S-box of excellent encryption properties is used.
The main attention of the attackers is to access the data being transferred via security
framework. Moreover, maliciously claiming the authentication and alteration of the data are
the core objectives of the hackers/aggressors. The visualization of design of block-cipher in
this regard is their imperative objective. Different guess of linear, differential and brute force
attacks may work for a weak cryptosystem, by bearing in mind some cipher-texts. The main
hurdle in block cipher is the mysterious look up table i.e. S-box. The stronger the nonlinear
component of block cipher, the more robust and secure is that cryptosystem.
Many physical systems are found chaotic in nature in the field of biology, physics and
engineering like weather forecast, movement of gases in atmosphere, stock exchange index
and similar events occurring in many areas of life. Chaos theory is utilized nowadays in
engineering, biology, physics and economics for the proper evaluation of dynamical systems.
Discrete systems are easy to model and can be predicted for even long periods unlike chaotic
dynamical systems. It is observed that every chaotic system is mathematically a nonlinear
8
system. The specific attributes of such systems are sensitivity towards initial
conditions/parameters, unpredictability, randomness, complex bifurcation pattern and
periodicity. These features make them challenging to analyse. Cryptography make use of such
features like unpredictability and randomness to design cryptosystem that canโt be predictable.
This thesis gives an application of chaotic dynamical systems in multimedia security,
specifically, these systems are utilized to build cryptographically strong S-boxes. Moreover,
these S-boxes are utilized in encryption schemes, watermarking techniques and steganography.
The only aim is to enhance the security of an encryption standard.
1.2 Research Objectives
The leading objectives of this research are as follow:
1. Identification of chaotic dynamical systems capable of producing rich and complex
dynamics based on mathematical schemes like ordinary differential equations.
2. Instead of using one dimensional systems for the production of one stream of
pseudorandom numbers, focus is to generate this target by using multi-dimensional
systems like coupled differential equations, for the production of more than two streams
of pseudorandom numbers.
3. Moreover, the all above mentioned mathematical systems are to be solved in such a
way that they can produce large number of S-boxes with different cryptographic
properties.
4. To establish efficacy of the above systems in applications of multimedia security,
techniques of image encryption and watermarking are to be developed.
5. Last objective is to achieve all above targets with low computational complexity.
Acknowledging the importance of nonlinear component of block cipher, the aim in this thesis
is to design new S-boxes having suitable/enhanced cryptographic properties. Furthermore, their
9
efficacy can be established by the applications in image encryption and watermarking schemes
by analysing the experimental results and observations. This can be achieved either by using
different mathematical structure/system or by applying several steps in construction to obtain
the required randomness. Our emphasis is to utilize chaotic dynamical systems for the
construction of S-boxes as well as encryption/watermarking schemes. Thus by this way we
expect to develop new schemes for the security of important stuff by utilizing known chaotic
systems.
1.3 Thesis Layout
The present theses is make up of total six chapters. The detail of all theses is given
hereafter.
Chaos and cryptography are mainly discussed in the current chapter. Some basic
terminology of cryptography and block ciphers are elaborated in detail. Moreover,
important properties of S-box in the perspective of mathematics are also explained.
The set of coupled differential equations i.e. Rabinovich-Fabrikant (RF) equations
generating chaotic behavior are initially utilized for the generation of three distinct
and unpredictable sequences of pseudorandom numbers in chapter 2. These
sequences are large enough to produce many chaotic S-boxes utilizing different
combination of the sequences. One simple and orthodox collection of pseudo random
numbers from a single sequence produced an S-box possessing fair enough
cryptographic traits making it capable of its further use in multimedia security.
The simplest chaotic dynamical system is double pendulum. Chapter 3 presents an
idea of designing nonlinear component of block cipher using double pendulum and
symmetric group. Symmetric group is used only for permutation at final stage. The
result analyses are promising for its further use in multimedia security.
10
An application of first order coupled differential equation in the form of image
encryption and watermarking is given in chapter four. Chaotic solution of
Rabinovich-Fabrikant (RF) system of equations with different initial conditions and
parameters is achieved using Runge-Kutta method of order four in MATLAB. Three
dissimilar sequences from the solution space are extracted for their further utilization
in color image encryption. Each channel of red, green and blue are separately
operated to scramble the original positioning of pixels to yield an encrypted
meaningless image. Moreover, the same sequences are implemented for
watermarking which was done along each channel. The simulation results of all major
standard analyses were observed carefully and found clear in determining the
robustness of the proposed schemes of encryption and watermarking.
In chapter five, a second order differential equation also known as Duffing equation
is debated. The chaotic behavior of this equation is seen in the form of oscillator. The
interesting attribute of this oscillator is the production of different bifurcation pattern
for slight variation in parameters/conditions causing chaos. Its density distribution
function is complex, rich and dense. It means one can generate a large number of S-
boxes from this single system. Lightweight but yet practical and applicable image
encryption based on chaotic differential equation is presented in this chapter. All the
standard analyses were found promising in analyzing the suggested scheme of
encryption.
This thesis has been ended with chapter 6 which includes the conclusions and the
future directions.
11
1.4 Cryptography and Chaos
In this section a very brief overview of chaotic system is given. Moreover, those
attributes of chaotic systems are explained which have somewhat influence in the design of
cryptosystems. Thus this section will highlight the relationship between chaos and
cryptography.
1.4.1 Chaotic System
Ordinarily, any physical system governed by some mathematical equations generating
dynamics that is unpredictable with the passage of time is known as chaotic system [9]. Chaos
is also known as confusion or disorder. Some system observing change over the time
sometimes gives the chaotic motion. Thus the change and time are the two basis of chaos
theory. The chaotic behaviour is detected by the graphical assessment of the time series of that
system. These systems donโt follow the trends, hence are unpredictable. There are numerous
naturally occurring and laboratory based planned dynamical systems (appearing in the fields
of engineering, electronics, physics, economics, ecology and many others) are found chaotic.
It is not necessary that chaotic behaviour will arise from a complex system of equations, rather
we can see chaotic motion from a simple equation as well [6], [7] and [8]. The only one variable
can appear in an equation generating chaos i.e. there is no particular restriction on number of
variables and parameters. Moreover, some systems are deterministic i.e. they follow a proper
set of equations or rules to predict the next term or region but still they are haphazard, such
system generates deterministic chaos. Furthermore, the chaotic motion of a system is self-
driven implying that no external involvement is required to produce chaotic solution. All these
lead to a difficult situation to identify chaos in real life problems. Although they are observed
in mathematics and computer science by graphic visualization of the governing
equations/problems.
12
The first and foremost attraction of scientist to chaos theory is the visualization of
complex and disordered behaviour of the system resulted from a simple deterministic equation.
Secondly, the system being under consideration is comprehendible at the same time impossible
to decipher and recognize from the solution trajectory. Third attraction lies in the minimum
background knowledge of progressive mathematics, one can understand the chaos with the
basic knowledge of algebra, geometry and calculus. Finally, chaos can be analysed without
going in depth of underlying mathematical equations. All these revelations surprises the
cryptographers and force them to make use of such systems to design strong cryptosystems that
are harder to decipher.
1.4.2 Chaotic Dynamical Systems
The chaos appears sometime, when a system observe change with the passage of time.
Temporal chaos and spatial chaos arises whenever time is replaced with the space and distance
respectively. Unlike linear systems, the nonlinear equations appearing in algebra or differential
equations are difficult to study. The dynamics of such systems is also complex. Furthermore,
every nonlinear system need not to be chaotic in nature. Many experts are of the view that
nonlinear dynamics or dynamical systems theory lies in the domain of chaotic dynamical
systems.
There are two categories of dynamical systems depending upon the conservation of energy.
Conservative dynamical system is one which does not loose energy i.e. friction free system.
While dissipative system is one which has to bear frictional forces and hence it loses energy.
Dissipative dynamical system after losing energy reaches to a limiting condition, under the
influence of certain constraints, give birth to a chaotic solution [9]. Our focus in this thesis
would be on chaotic dynamical systems hence dissipative dynamical systems in particular.
The change appearing in a dynamical system can be observed in discrete intervals of time.
These intervals of time may be evenly distributed or unevenly. Examples of such systems
13
includes rainstorms, earthquakes and volcanic explosions. Discrete time systems are governed
by the difference equations that are solved iteratively. The variations of a dynamical systems
are also observed for continuous intervals of time. There is a continuity in the measurement of
such phenomenon unlike discrete intervals of time. The differential equations are used for the
measurement of continuous change of a dynamical system. Examples of such systems includes
the air temperature, heat conduction and water flow in rivers [9].
Differential equations being vast and developed branch of mathematics, are found in almost all
fields, wherever a physical system is modelled. The comprehensive grip of mathematicians on
this field have managed to forecast various phenomena in acoustics, astrophysics, weather
updates and many other life sciences. The idea is to implement differential equations in cyber-
security for the design of secure and robust systems. The only nonlinear component of block
cipher is designed with the help of nonlinear systems of differential equation that can create
hurdles for cryptanalysis.
1.4.3 Causes of a Chaotic System
Chaos theory is accurately a multidiscipline topic. The importance of chaos in the recent
decades has been recognized by many scientists by considering such systems for their proper
evaluation and examination. The factors causing chaos in real world phenomenon are still
unknown. To some extant one can say that the factors causing chaos are variations in control
parameters, deviations of initial conditions, nonlinear interaction of two or more progressions,
involvement of nonlinear terms in the equations and noise/resistance.
1.4.4 Characteristics of a Chaotic System
The peculiar nature of a chaotic dynamical system is still a well-known problem for the
scientists. With the advancement of computing devices, although the bifurcation pattern of
chaotic systems are visualized by using different software but still the proper grip on the this
14
subject is an objective of many. The unusual behaviour is observed by all in analysing these
systems[9]. The specific attributes of a chaotic system are as follow.
Sensitivity to Initial Conditions
The most important property of a chaotic system is its sensitive behaviour towards initial
conditions and parameters. Sometimes slight variations in initial input results totally different
bifurcation pattern of that system. Cryptographically speaking, this property is most attractive
for the design of a cryptosystem. It assures the sender that any slightly wrong guess will
generate a different solution space, hence predictability of data is minimized.
Entropy
The amount of disorder is usually evaluated in entropy analysis. Since chaotic system bears the
most disorderly behaviour. Thatโs why entropy is linked with these systems.
Lyapunov Exponent
Lyapunov exponent is used to decide whether the mathematical system used is chaotic or not.
The value of this exponent greater than zero implies the chaotic nature of a system.
Long Term Unpredictability
This characteristic of chaotic system is very interesting and irrational. The bifurcation pattern
and trajectory of such system is unpredictable for very long interval. This is very useful for the
utilization of such systems in cyber security.
1.4.5 One Dimensional Discrete Chaotic System
The chaotic system used in cryptography are reviewed hereafter on the basis of dimensions
initially. Mostly, Simple chaotic systems having one dimension are used in cryptography that
are easy to understand and evaluate as compared to complex systems. The shortcoming lies in
these are like small solution space, can be predicted using advanced technology, fewer number
of controlled parameters and conditions and smaller key space. Due to all these the recapture
of such system using different software is comparatively an easy task. [9].
15
The Logistic Equation
The population model in terms of an equation known as logistic equation was proposed by a
biologist Robert May in 1976 [10]. This is the simplest discrete time intervals based chaotic
system. It explains various key features of a chaotic system. Moreover, many researchers have
used this map because of its simplicity in cryptography [8] and [11]. Mathematically, its map
is represented as
๐ฅ๐ +1 = ๐พ๐ฅ๐ (1 โ ๐ฅ๐ )
Tent Map
Tent map is an iterative map generating deterministic chaos under certain particular selection
of the parameter ๐, which is considered responsible for controlling chaos. The range of ๐ for
this map is in [1, 2] for chaotic behaviour.
๐ฅ๐ก+1 = 2๐ฅ๐ก ๐๐ 0 โค ๐ฅ โค 0.5
๐ฅ๐ก+1 = 2 โ 2๐ฅ๐ก ๐๐ 0.5 < ๐ฅ โค 1
This map with slight extension in it has been utilized by [8] in the field of chaotic cryptography.
Quadratic Map
The term quadratic refers to the polynomial of degree 2. The standard quadratic equation is
๐๐ฅ2 + ๐๐ฅ + ๐ = 0
The constants a, b and c are to define chaotic behaviour for this quadratic map. For example
๐ฆ๐+1 = ๐ + ๐ฆ๐2
Where c is a constant. The above equation is a special case of standard quadratic equation for
the case a=1 and b=0. A similar quadratic equation generating chaos is of the type
๐ค๐ก+1 = (๐ค๐ก โ 2)2
Gingerbreadman Map
It is a two dimensional chaotic map presented initially by [12], mathematically it is defined by
a piecewise linear transformation as follow:
16
๐ฅ๐+1 = 1 โ ๐ฆ๐ + |๐ฅ๐|
๐ฆ๐+1 = ๐ฅ๐
[13] used this map in the design of cryptosystem.
Henon Map
This is an example of discrete time dynamical system proposed by Michel Henon [14] in
explaining the Poincare section of the Lorenz model. It involves two constants that control
chaos. Mathematically,
๐ฅ๐+1 = 1 โ ๐ฆ๐+1 โ ๐๐ฅ๐2
๐ฆ๐+1 = ๐๐ฅ๐
[15] Utilized this map for the construction of S-box.
Other Chaotic Maps
There are some other chaotic maps available in literature that exhibit chaotic behaviour and
can be applied for the design of cryptosystem [9].
Climatology
๐ง๐ก+1 = ๐(3๐ง๐ก โ 4๐ง๐ก3)
Biology
๐ข๐ +1 = ๐ข๐ ๐๐(1โ๐ข๐ )
Various Physical System (Thomas and Stewart)
โ๐+1=โ๐+๐
2๐sin(2๐ โ๐) + ๐
Various Physical System (Jensen)
๐ฅ๐+1 = ๐ฅ๐ + ๐ฆ๐+1
๐ฆ๐+1 = ๐ฆ๐ + asin (๐ฆ๐)
17
1.5 Differential Equations Based Chaotic Dynamical System
The examples from a continuous category i.e. differential equations are listed here. These are
actually the necessary helping tools for the better understanding of chaotic dynamical systems.
1.5.1 Rossler Attractor
In the year 1976, Rossler presented the following coupled ordinary differential equations for
the generation of chaotic behaviour [16]
๐๐ฅ
๐๐ก= โ๐ฆ โ ๐ง
๐๐ฆ
๐๐ก= ๐ฅ + ๐๐ฆ
๐๐ง
๐๐ก= ๐ + ๐ง(๐ฅ โ ๐)
1.5.2 Lorenz System
In the study of fluid flow under the branch of climatology, Lorenz developed the following set
of differential equations in the year 1963 [17],
๐๐ฅ
๐๐ก= โ๐ผ โ ๐ผ๐ฆ
๐๐ฅ
๐๐ก= โ๐ฅ๐ง + ๐ฝ๐ฅ โ ๐ฆ
๐๐ฅ
๐๐ก= ๐ฅ๐ฆ โ ๐พ๐ง
Where ๐ผ, ๐ฝ, ๐๐๐ ๐พ constants.
1.5.3 Rabinovich-Fabrikant System of Equations
This set of coupled differential equations are presented in 1979 by Mikhail Rabinovich and
Anatoly Fabrikant [18]. This system is of chaotic nature relying upon three variables and two
18
parameters. It models the uncertainty arising from the modulation instability in a non-
equilibrium dissipative medium. The set of equations are as follow
๐๐ฅ
๐๐ก= ๐ฆ๐ง โ ๐ฆ + ๐ฆ3 + ๐๐ฅ
๐๐ฆ
๐๐ก= 3๐ฅ๐ง + ๐ฅ โ ๐ฅ3 + ๐๐ฆ
๐๐ง
๐๐ก= โ2๐ฝ๐ง โ 2๐ฅ๐ฆ๐ง
For different parameters and initial conditions the system possess unalike bifurcation pattern
confirming its sensitivity towards initial input.
1.5.4 Double Pendulum
One of the simplest chaotic dynamical system is double pendulum. It is formed by joining two
simple pendulums from end to end. The movement behaviour is dependent upon the inclination
of the pendulums i.e. for a very minute change in initial angle yields totally different bifurcation
behaviour. This gets more interesting in terms of complexity and chaos when two double
pendulums are moved simultaneously in a single pivot. Their mathematical representation
comes in terms of four first order differential equations.
1.5.5 Second order Differential Equation (Duffing Oscillator)
This chaotic oscillator originates from the solution of second order nonlinear differential
equation. It models certain damped and driven oscillators [19]. For a displacement vector ๐ฅ,
the first and second derivative of ๐ฅ denotes the velocity vector and acceleration respectively.
Mathematically duffing equation is,
๐2๐ฅ
๐๐ก2+ ๐ผ
๐๐ฅ
๐๐ก+ ๐ฝ๐ฅ + ๐พ๐ฅ3 = ๐cos (๐๐ก)
19
Mathematical Systems
Linear Systems
Discrete
Chaotic System
Nonlinear Systems
Chaotic Systems
Ordinary Differential
Equation
2nd and Higher
order ODE
Coupled ODEs
Continuous
Chaotic System
Partial Differential
Equation
Iterative System
Logistic map, Tent map,
Chebyshev map etc..
1st Order ODE
Fig. [1]. Flow Chart of origin of chaotic system from mathematical equation
1.6 Boolean Algebra
It is necessary to understand Boolean function and algebra before going in depth of soft
computing devices. This is branch of mathematics which has contracted the whole real line into
two output i.e. zero and one. The invention of microprocessor and speedy systems is due to
this contracted real line. It is also mandatory to study it for better understanding of block ciphers
and S-box. Moreover, the information transmitted is also converted into bits and bytes for
further processing that also require pre knowledge of Boolean algebra. A very brief and
fundamental definitions connected to this thesis are discussed below.
20
1.6.1 Boolean Function
If ๐บ๐น(2๐) is a n-dimensional vector space, Boolean function in terms of mapping is defined as
๐(๐ฅ) โถ ๐บ๐น(2๐) โ ๐บ๐น(2)
Here, ๐บ๐น(2๐) is the Galois field having 2๐ elements in a binary form and ๐ฅ = (๐ฅ1, ๐ฅ2, โฆ , ๐ฅ๐).
The total number 22๐
of unique Boolean function from this representation can be assembled.
The scalar product of the two vectors ๐ฅ and ๐ฆ in ๐บ๐น(2๐) is defined as
< ๐ฅ, ๐ฆ > = โ๐=1๐ ๐ฅ๐ . ๐ฆ๐
These functions can be expressed with the help of polarity truth table and truth table.
1.6.2 Hamming Weight
It is one of the basic term which is further used in explaining many terminologies related to the
theory of the Boolean functions. Consider a Boolean function ๐(๐ฅ) of n variables, the number
of ones (1s) in the truth table are known as hamming weight of Boolean function [20].
โ(๐ค) = 2๐โ1 โ1
2โ๏ฟฝฬ๏ฟฝ(๐ฅ)
๐ฅ
=โ๐(๐ฅ)
๐ฅ
1.6.3 Hamming Distance
Consider two functions ๐, โ ๐๐2๐ then hamming distance ๐(๐, โ) is the number of truth table
positions which are different with each other [20]. It is also known as hamming weight of the
XOR sum of two functions
๐(๐, โ) = โ(๐ โ โ)
The similarity between โ ๐๐๐ ๐ is characterized by the use of hamming distance. This
resemblance is linked with the idea of correlation of two functions that is critical to analyse
from a cryptographic perspective. The correlation coefficient of totally uncorrelated functions
is zero whereas it is one for the totally correlated i.e. โ = ๐.
21
1.6.4 Correlation
The correlation between ๐ ๐๐๐ โ is defined [20] as follow
๐ถ๐๐๐(๐, โ) = 2๐(๐ = โ) โ 1
= 2 [2๐ โ ๐(๐, โ)
2๐] โ 1
2 (1 โ๐(๐, โ)
2๐) โ 1
๐ถ๐๐๐(๐, โ) = 1 โ๐(๐, โ)
2๐โ1
Using definition of ๐(๐, โ)
๐ถ๐๐๐(๐, โ) = 1 โโ (๐(๐ฅ)โ โ(๐ฅ))๐ฅ
2๐โ1
=โ [1 โ 2(๐(๐ฅ) โ โ(๐ฅ))]๐ฅ
2๐
=โ (โฬ(๐ฅ)๏ฟฝฬ๏ฟฝ(๐ฅ))๐ฅ
2๐
The outcomes of ๐ถ๐๐๐(๐, โ) lies in the interval [โ1,1]. The hamming distance between two
functions defines the ๐ถ๐๐๐(๐, โ) i.e. 1 implying hamming distance zero and -1 is for the
maximum hamming distance 2๐. For a pair of a function, ๐ถ๐๐๐ is an important component
defining imbalance in functions.
1.6.5 Algebraic Normal Form (ANF)
Algebraic normal form (ANF) is another way of expression for a Boolean function [20]. This
utilizes the idea of AND products of inputs with XOR sum. For every ANF representation there
exist unique truth tale of a Boolean function. If ๐(๐ฅ) contains all n-variables in ANF
representation then it is called as nondegenerate function otherwise it is degenerate.
The algebraic degree of the function is deg (๐) defined as number of variables in the largest
product term having nonzero coefficient in ANF representation of ๐(๐ฅ).
22
1.6.6 Walsh Hadmard Transform
Another way of defining the Boolean functions is with the help of Walsh Hadmard transform
(WHT). It measures the correlation between function and its set of linear functions.
๐(๐ข) =โ(โ1)๐(๐ฅ)(โ1)๐๐ข(๐ฅ)
๐ฅ
The WHT value for different Boolean function is unique.
1.7 Cryptographic Properties of Substitution box
The only nonlinear element in block ciphers is the substitution box (S-box). To support the
research conducted in this thesis, some very basic and fundamental explanation of theory of S-
box is presented here. Moreover, cryptographically the important features of S-box are also
given here.
1.7.1 S-Box
The customary development of the theory of single input to numerous output is dealt by S-box.
An S-box of dimension ๐ ร ๐ is a nonlinear mapping operating on ๐ input bits and generating
๐ output bits.
๐: ๐2๐ โ ๐2
๐
1.7.2 Nonlinearity
In order to resist linear attacks against the approximation of an S-box, the extent of nonlinearity
is crucial. Higher values of nonlinearity implies robustness and stiffness against linear
approximation. Consider a set ๐ containing all possible linear combinations of the columns of
an S-box ๐ which are all nontrivial, then nonlinearity NL(T) = min๐๐ฟ(๐), where ๐ โ ๐. In
other words, the minimum probable hamming distance in between the S-boxโs component and
23
all affine functions of n variables gives the nonlinearity [21]. In other words, Nonlinearity is
the smallest hamming distance of a Boolean function to the collection of affine functions [22].
๐๐ฟ(๐) = 2๐โ1 โ1
2|๐๐๐๐ฅ|
1.7.2 Bit Independence Criterion
As the name of the analysis bit independence criterion (BIC) indicates that for any change in
bits makes the other bits to change independently. If for 2 2: m mh Z Z , we have ๐ฅ, ๐ฆ, ๐ง โ
{1,2,3, โฆ ,๐} and ๐ฆ โ ๐ง, then any small or big change in ๐ฅ, results in variation of ๐ฆ ๐๐๐ ๐ง
output bits autonomously, then BIC is satisfied [22]. BIC studies the effect of complementing
of a single input bit on whole output bits. Hence, independent behaviour of two avalanche
vectors in pairs and the variation of input bits are imperative factors for BIC. The range values
for BIC are in [0, 1], where lower value of this interval is considered as an ideal one and upper
value appears to be the worst.
1.7.3 Strict Avalanche Criterion (SAC)
In any substitution-permutation network, Avalanche effect is observable whenever a chain of
deviations are produced due to the consequence of a solitary input disparity [20]. For a function
๐: ๐2๐ โ ๐2
๐, SAC is satisfied if the following holds
โ ๐(๐ข)โจ๐(๐ขโจ๐ถ๐๐)
๐ขโ๐น2๐
= (2๐โ1, 2๐โ1, โฆ , 2๐โ1)
This criteria is satisfied for every strong Boolean S-box along with the completeness
property. A single input bit influences half of the total output bits, is indeed an interesting
observation. This is basically inferring strong resistance against the plaintext attacks. If it
doesnโt happen then an attacker can predict the pattern by observing outputs for different
24
inputs. SAC was coined by [23] by amalgamating the completeness and avalanche property of
a Boolean function. The acceptable probability value for SAC is 1
2 [8].
1.7.4 Linear Approximation Probability
The maximum amount of imbalance of an event is measured in the linear approximation
probability (LAP) [24]. It is defined mathematically as follow:
๐ฟ๐ด๐ = max๐๐,๐๐โ 0
|#{๐ฃ/๐ฃ. ๐ฃ = ๐(๐ฃ). ๐๐ }
2๐โ1
2|
Where 2๐ is the number of all possible inputs for a set ๐ธ consisting of all possible inputs and ๐๐,
๐๐ are representing input and output values respectively.
1.7.5 Differential Approximation Probability
The differential uniformity exhibited by a substitution box is measured with the help of
differential approximation probability. In this every input bit is closely examined to assure that
unvarying mapping is used to design the S-box.
# ( ) ( )
2f k
u U f u f u u zDP
1.7.6 Majority Logic Criterion
In this criterion, an image encrypted by an S-box has to satisfy majority logic criterion MLC.
It involves the analyses like homogeneity, energy, contrast, correlation and entropy analyses
[25]. These all analyses are used to evaluate the changes occurred in an encrypted image. In
other words, these are used to measure the strength of an S-box for image encryption. Detailed
discussion about MLC is given below in different chapters wherever it is applied.
25
1.8 Randomness Test (NIST SP-800 22)
The randomness test authenticated by national institute of standards (NIST) also known as
NIST SP 800-22 [26] is used to check statistically randomness and unpredictability of
pseudorandom number generators for cryptographic applications. This contains fifteen tests
including frequency (mono-bit) test, block frequency, long runs of ones, overlapping templates,
approximate entropy and random excursions etc. The basic fundamental test among these is
the frequency test which measures the number of ones and zeroes in a sequence. The remaining
fourteen tests depend mostly upon the passing of this test. If this fails than the chance of failures
of others increases. Thereafter, the test suit software evaluates the remaining tests to declare
the randomness of input data.
26
Chapter 2
First Order Coupled Differential Equations based
Dynamical System for the Construction of S-
boxes
In this chapter, first order coupled ordinary differential equations perceiving third order
nonlinearity generating rich chaotic and complex dynamics are utilized in cyber security.
Initially, this system will be solved to generate random integers, then for the production of
chaotic substitution boxes these integer values are permuted for obtaining highly nonlinear
chaotic S-box. The prime advantage of the proposed design is the construction of different
cryptographically strong S-boxes, by slightly altering the parameters and initial conditions of
the system of differential equations. An S-box constructed by utilizing this scheme is evaluated
by the algebraic and statistical analyses already available in literature. The outcomes of analysis
yielded promising statistics which ensure its importance in application of secure
communications.
2.1 Background
The communication over the globe via different devices is becoming a necessity of mankind
and is increasing exponentially with the passage of time. It includes the sensitive as well as
confidential data from the fields of medicine, engineering, foreign ministries and military etc.
There is always a threat of data leakage/modification while communication over the internet.
Many organizations are spending huge amount of sum for the safe communication of original
27
data. The data being transferred on internet is of delicate nature and cryptanalyst sitting in the
path anticipate illegitimate interception, manipulation and illegal usage of secret information
which is to be protected. For the safe transmission of valuable information researchers make
use of cryptographic algorithm.
Cryptography is an art of hiding data from initial stage to final stage by means of some
cryptographic algorithms so that no one in the path of communication can extract the vital
information transmitted. The only authentic users will be having access to the processed
information, if provided by suitable algorithm and keys by the sender. Symmetric and
asymmetric are the two main types of cryptosystems. Symmetric cryptography involves the
same key for encryption and decryption, while asymmetric cryptography is opposite to this.
Symmetric cryptography is further divided into stream ciphers and block ciphers [20]. Block
cryptosystem is the motivation behind this chapter. In block cryptograms the plaintext is
divided into blocks for the step by step utilization of cryptographic algorithm.
Shannon [1] introduced the two main themes of block cryptosystem i.e. confusion and
diffusion. Block cryptosystem involves four steps i.e. substitution, permutation, mixing and
adding key [21], [22], [27], [28] and [29]. The algorithms of block cryptosystem initially spread
the primary data into blocks of the comparable size thereafter encryption is done for the whole
block. Diffusion is the procedure in which the plaintext containing original message of the
sender is altered and made nebulous by scattering the original text bits to the cipher text bits.
Whereas, confusion is the procedure in which changing the original text changes the cipher
text. These two properties are generally acquired by means of round recurrence.
Differential equations (ordinary or partial) are significantly used to envisage the world
(surrounding) around us, which includes all fields from biology to engineering. A physical
problem is modelled in terms of differential equations, involving derivatives of an unknown
function to forecast the behaviour of the physical system in near future. Some of the equations
28
are classified as nonlinear differential equations and among these very few become chaotic
differential equations, i.e. there are certain systems (problems) that generate chaos under the
constraint of certain parameters. Such chaotic dynamical systems are extremely sensitive
towards the initial conditions and observe the behaviour of randomness, hence depicts different
trajectories for different initial conditions. Due to this behaviour of randomness, such chaotic
dynamical systems are utilized to create confusion and diffusion in cryptosystem.
The motivation behind the utilization of chaotic differential equations in block cryptograms is
the inherited unpredictability property of a chaotic system. There is a close connection among
the chaotic systems and the specific attributes of block cryptosystems like confusion and
diffusion. This relationship is due to the fact of sensitivity of the system towards initial
conditions/parameters, randomness, complex and chaotic behaviour of the system. Moreover,
the larger solution space tolerates the numerous schemes of robust and secure S-boxes from a
single system. This idea is different from the existing schemes due to the involvement of a
three dimensional system instead of one dimensional systems like logistic [8], tent [30] and
Chebyshev maps [31]. Furthermore, third order or a cubic nonlinearity of RF system as
compared to quadratic nonlinearity of Lorenz system [32] is the prominent feature responsible
for complex and chaotic dynamics for the proposed scheme and hence best suited for the design
of an S-box.
2.2 Coupled Differential Equations Based Construction of S-box
Rabinovich-Fabricant (RF) is a chaotic system [18] comprising of three coupled ordinary
differential equations with three variables and two fixed parameters as given below
๐๐ฅ
๐๐ก= ๐ฆ๐ง โ ๐ฆ + ๐ฆ3 + ๐๐ฅ
๐๐ฆ
๐๐ก= 3๐ฅ๐ง + ๐ฅ โ ๐ฅ3 + ๐๐ฆ (2.1)
29
๐๐ง
๐๐ก= โ2๐ฝ๐ง โ 2๐ฅ๐ฆ๐ง
This scheme models the randomness arising from the modulation instability in a non-
equilibrium dissipative medium. This was the work of Mikhail Rabinovich and Anatoly
Fabrikant [18] in the late eightyโs. Rabinovich-Fabrikant system is of highly chaotic nature and
sensitive to initial conditions, hence best suited for the design of block ciphers nonlinear
component. The chaotic behaviour of this system is observed even for the small change in
parameter and initial conditions. Fixation of initial conditions and parameters at one stage can
generate an S-box of encompassing best suited cryptographic properties. Some variations in
initial conditions and parameters yielded the following different chaotic trajectories as shown
in Fig. 1 and 2.
Figure 1. Different chaotic behaviour of RF system when only ๐ฝ is changed
Figure 2. RF systemโs behaviour by slightly altering initial conditions
RF system observes complex dynamics due to the third order nonlinearity involved in it. This
type of behaviour of coupled differential equations is rarely seen. Due to this, the analysis of
hetroclinic or homoclinic orbits, existence of invariant sets and so on is exceptionally difficult
to investigate. Even the available software packages that utilize numerical methods for the
-2
-1.5
-1
-0.5
0
-1
0
1
2
30
0.5
1
1.5
Rabinovich Fabrikant for sigma=0.87, beta=1.1
-1.4-1.3
-1.2-1.1
-1-0.9
-0.5
0
0.5
1
1.50
0.1
0.2
0.3
0.4
0.5
Rabinovich Fabrikant system for sigma=0.1, beta=0.98
-3
-2
-1
0
1
-4
-2
0
2
40
0.5
1
1.5
2
Rabinovich Fabrikant System for sigma=0.1, beta=0.14
-1.65-1.6
-1.55-1.5
-1.45-1.4
-0.1
0
0.1
0.2
0.3
0.35
0.4
0.45
0.5
0.55
-4
-2
0
2
4
-4
-2
0
2
40
0.5
1
1.5
2
-6-4
-20
24
6
x 106
-6-4
-20
24
6
x 106
0
0.5
1
1.5
2
x 1011
30
solution of differential equations will yield different dynamics for the same initial conditions
and parameters [33].
Substitution box plays a crucial role in accomplishing outstanding cipher properties. Its
importance in any cryptosystem cannot be denied at all because of its resistance against
differential and linear cryptanalysis. In any substitution permutation network, an S-box is
responsible in creating confusion and disarray in cryptosystem. Mathematically, the
substitution box is defined as:
S: โค2๐ โ โค2
๐.
The Lorenz system used by Majid et.al [32] contains the nonlinearity of second order. The
system of equations utilized here is much better than preceding one because it contains a third
order nonlinearity generating strong complex and chaotic dynamics. The range of parameters
for the chaotic bifurcation pattern as observed on MATLAB is:
Fix beta at 1.1, choose sigma lying in between [0.13, 0.87] for chaotic solution.
Fix sigma at 0.87, choose value of beta greater than 0.1 for chaotic solution.
Instead of variation of parameters one can also obtain the chaotic bifurcation pattern of
the system by slightly altering the initial conditions.
The steps involved in the construction of substitution box are explained as follow:
Initially, for the solution of system of equations (1), we have fixed sigma at 0.87 and
beta at 1.1 and initial conditions are chosen as (-1, 0, 0.5).
Numerical solution is obtained in MATLAB using Runge-Kutta of 4th order.
In this step extract ๐ฆ from the [๐ฅ, ๐ฆ, ๐ง] solution of system of equations (1) and name this
sequence as ๐บ.
Multiplying ๐บ with 1000 to get a new sequence of numbers ๐พ
๐พ = ๐บ โ 1000
The sequence of numbers ๐ป is obtained using ๐๐๐๐(๐๐๐(๐1 โ 256,256)).
31
๐ป = ๐๐๐๐(๐๐๐(๐พ โ 256,256)) (2.2)
In the last step, the sequence obtained above is permuted in MATLAB to obtain an S-
box of desirable cryptographic properties.
All steps are explained in the flowchart given in Fig.3.
Start
g1 = y ร 1000
Fix Parameters and initial
coordinates to obtain numerical
solution in MATLAB
Extract y from the solution
[x,y,z]
Arrange into 16 ร 16
matrix to obtain S-boxCeil(mod((g1ร256, 256)) End
Fig. 3: Flow chart for the construction of S-box
Table 1: Designed S-box
123 131 87 199 57 165 157 47 172 132 205 139 203 116 97 146
162 149 81 77 183 5 202 108 55 4 75 128 170 174 41 222
148 10 155 62 15 63 229 201 140 53 67 236 215 107 100 136
239 32 211 69 198 89 42 35 160 135 30 184 210 253 96 25
252 232 235 43 61 247 34 228 185 204 85 141 158 175 194 111
191 200 238 245 147 251 23 152 180 13 168 143 193 6 125 117
83 197 177 106 98 79 166 224 231 169 50 218 178 124 119 206
130 192 227 48 219 16 86 241 74 233 60 88 31 84 8 156
94 54 9 12 186 28 250 196 37 65 151 246 213 93 52 225
255 101 137 127 92 216 189 38 26 164 159 33 248 36 113 45
190 214 17 134 40 59 20 72 22 120 110 44 179 68 0 161
249 64 80 3 51 223 105 114 19 118 29 181 82 244 230 91
243 2 39 121 240 11 58 182 109 90 122 221 173 78 138 209
167 70 187 171 14 234 226 126 27 49 95 66 7 76 154 115
32
46 153 133 150 254 163 208 99 73 24 129 176 112 18 212 71
237 102 242 21 207 103 144 1 220 145 104 195 188 217 142 56
2.3 Algebraic and Statistical Result Analyses of S-box
This section is used to evaluate the efficiency of the block ciphers nonlinear element. The
analyses that establish its efficacy includes nonlinearity analysis, bit independence criterion
(BIC), strict avalanche criterion (SAC), linear and differential approximation probabilities. The
details of these analyses are given henceforth.
2.3.1 Nonlinearity
Among all the cryptographic properties the nonlinearity is the most imperative one. For a strong
cryptographic system, it is necessary that nonlinearity must be higher. Mathematically, if we
represent nonlinearity by ๐๐ฟ๐ , then it is defined as
๐๐ฟ๐ =1
2(2๐ โ๐๐ป๐๐๐๐ฅ) (2.3)
Where ๐๐ป๐๐๐๐ฅ represents maximum absolute value of Walsh Hadamard transform vector.
Which calculates the resistance of a system being expressed as a set of linear equations and
hence confirms resistance against linear cryptanalysis.
Table 2: Nonlinearity comparison table for proposed S-box
S-boxes Proposed Ref. [32] Ref. [8] Ref. [34] Ref. [35] Ref. [36] Ref. [24]
Average 107.5 104.7 106.75 103 103.3 105.25 108.88
Minimum 106 102 106 100 99 102 103
Maximu
m
110 108 108 106 106 108 112
2.3.2 Bit Independence Criterion
To fortify the proficiency of confusion function in substitution boxes, the idea of bit
independent criterion (BIC) is utilized. This statistical property was firstly defined by Webster
33
and Tavares [37], i.e. for a collection of certain avalanche vectors; overall the avalanche
variables have to be pairwise independent.
Table 3: Comparison table of BIC values for various S-boxes
S-boxes Proposed Ref. [32] Ref.[8] Ref.[34] Ref.[35] Ref.[36] Ref. [24]
BIC-SAC 0.5048 0.5058 0.4989 0.5050 0.4987 0.4956 0.5052
BIC
Nonlinearity
105 104.1 106.6 103.1 103.3 103.8 107.46
2.3.3 Strict Avalanche Criterion
Webster and Tavares in the year 1985 introduced the concept of strict avalanche effect (SAC)
which is the generalization of completeness and avalanche effect. SAC is delineated as all the
output bits differ by a probability of a half whenever a single input bit is complemented.
โ: ๐บ๐น(2)๐ โ ๐บ๐น(2)๐
๐๐๐๐(โ(๐ฅ๐))๐ โ ๐๐๐๐(๐(๐ฅ))๐ =1
2 โ ๐ โ [1, ๐] ๐๐๐ ๐ โ [1,๐] (2.4)
Table 4: SAC values Comparison table for various S-boxes
S-box Propose
d
Ref. [32] Ref. [8] Ref. [34] Ref. [35] Ref. [36] Ref. [24]
Minimum 0.4065 0.3906 0.4219 0.4218 0.4140 0.4297 0.3907
Average 0.500 0.506 0.4939 0.500 0.499 0.496 0.502
Maximum 0.5785 0.5937 0.5625 0.6093 0.6015 0.5313 0.6133
2.3.4 Linear Approximation Probability
The linear approximation probability (LAP) is in fact the largest value of the disparity of an
event. Let ๐ be set of all probable inputs with total elements 2๐. If input and output
concealments are ๐๐ ๐๐๐ ๐๐ respectively, then (LAP) values, given in Table 5, are obtained by
the following equation:
34
๐ฟ๐ด๐ = max๐๐,๐๐โ 0
|#{๐ข/๐ข. ๐ข = ๐(๐ข). ๐๐ }
2๐โ1
2| (2.5)
Table 5: Comparison table of LAP values for different S-boxes
S-box Propose
d
Ref. [32] Ref. [8] Ref. [34] Ref. [35] Ref. [36] Ref. [24]
Max.
LAP
0.148 0.1250 0.1250 0.1289 0.1328 0.1562 0.128
Max.
Value
164 160 160 162 164 168 162
2.3.5 Differential Approximation Probability
The differential approximation probability also known as differential homogeneity is
considered as an ultimate trait of an S box. The degree of a differential equality, also known as
differential approximation probability is defined as follow:
๐ท๐ = (#{๐ฃ โ ๐ โง ๐(๐ฃ)โจ๐(๐ฃโจโ๐ฃ) = โ๐ง}
2๐) (2.6)
Table 6: Differential approximation probability of proposed S-box
S-box Proposed Ref. [32] Ref. [8] Ref. [34] Ref. [35] Ref. [36] Ref. [24]
Max. DP 0.03625 0.04688 0.0625 0.05469 0.03906 0.03906 0.0312
2.3.6 Majority Logic Criteria
Majority logic criterion (MLC) contains the comprehensive debate about correlation analysis,
homogeneity, contrast, energy, and entropy analysis [25]. For the purpose of image encryption,
an S-box must have a good statistical strength which is weighed in MLC analysis. Due to
encryption algorithm, the change appearing in original image after encryption is the source of
measuring the strength of an S-box. Some very brief details of these analyses are discussed
below and pictorial representation is given figure 7 and 8.
35
Information entropy is used to measure the statistical amount of disorder in an encrypted image.
If the total number of pixels in image is ๐, then measure of the information entropy of a
ciphered image is:
E(U) =โ๐(๐ข๐) log2 ๐(๐ข๐)
๐
๐=1
(2.7)
Where, ๐(๐ข๐) is the probability of the pixel ๐ข๐ . The theoretical value of entropy is 8 for a grey
scale image, when the chance of pixel occurrence is same. Therefore to validate the efficacy of
the proposed scheme, the information entropy must be approaching to 8. Table 7 gives the
comparison of entropy analysis of the proposed scheme with [27], [8] and [36].
The adjacent pixels (horizontally, vertically and diagonally) of the original image are extremely
correlated. Encryption procedures are used to make these pixels unrelated. The correlation of
adjoining pixels of an enciphered image approaching to zero indicates that the scheme is secure
and robust. The following formula is used to calculate the correlation between adjacent pixels
๐ ๐๐๐ ๐ of the grey level image
ฮ๐๐ =โ ({๐๐ โ ๏ฟฝฬ ๏ฟฝ}{๐ ๐ โ ๏ฟฝฬ ๏ฟฝ})๐๐=1
โโ (๐๐ โ ๏ฟฝฬ ๏ฟฝ)2๐๐=1 โโ (๐ ๐ โ ๏ฟฝฬ ๏ฟฝ)2
๐๐=1
(2.8)
The outcomes produced by above equation are presented in table 7. The correlation coefficient
of the plaintext image and enciphered image is nearly equal to one and zero respectively as
desired.
The loss of brightness from the plaintext image after encryption is evaluated in contrast
analysis. The higher amount of contrast for the encrypted image is due to the secure encryption
scheme. Moreover, energy and homogeneity analyses are used to measure the characteristics
and behaviour of an encrypted image.
Table 7: Majority logic criterion MLC for numerous S-boxes
Pictures Entropy Correlation Energy Homogeneity Contrast
36
Original 7.1278 0.6849 0.0895 0.7488 0.8566
Proposed S-box 7.9775 0.0006 0.0158 0.3905 10.5427
Ref. [27] 7.9591 -0.0441 0.0202 0.4151 8.2314
Ref. [8] 7.9812 -0.0045 0.0177 0.4091 8.3154
Ref. [36] 7.9431 0.0155 0.0219 0.4248 8.2113
Ref. [2] 7.9561 0.0554 0.0202 0.4662 8.3124
Figure 4. Original image and the encrypted image of Baboon using the proposed S-box
The pixels of an image are initially converted into binary string of eight bits. The first
four bits occurring at the left position specify the row of an S-box whereas the right four bits
determine the column of an S-box. The entry at the intersection of the suggested row and
column of an S-box substitute the pixel value of an image. This procedure goes on until the
whole pixels are substituted by the S-box entries. This procedure generates an enciphered
image. The host and corresponding encrypted images are shown in Fig. 4(a) and 4(b)
respectively. Their corresponding histograms are given in Fig. 5. The result analyses of MLC
are also promising and suggesting the suitability of the proposed S-box in the design of block
ciphers for the secure communication of data.
37
Figure 5. Histogram for original image and the encrypted image of Baboon
The need of developing a new encryption standard after the discovery of advanced encryption
standard (AES) is minimized. It is the ultimate desire of researchers to design cryptographically
strong substitution boxes that are used in symmetric key cryptography. Keeping this idea in
mind, a novel method based on Rabinovich Fabrikant system of equations for construction of
S-box is presented. Third order nonlinearity of RF system is responsible for chaotic and
complex dynamics. Additionally, this system is highly sensitive to initial conditions and
parameters hence best suited for the design of nonlinear component of block ciphers. The
available literature material related to the presented scheme is Lorenz system [32], [38], which
contains the nonlinearity of second order, but RF system contains a nonlinear term of third
order. This idea is utilized for the first time in substitution-permutation like network. The result
analyses of the proposed S-box are far better than chaotic S-boxes, which confirms its efficacy
in practical application of cyber security. It is also suggested that there is no need of making
different rounds while encrypting an image using this chaotic dynamical systems because
single round will probably yield the desired results due to the complex and chaotic nature of
the system.
0
100
200
300
400
500
600
700
800
900
0 50 100 150 200 250
0
100
200
300
400
500
600
0 50 100 150 200 250
38
Chapter 3
Construction of Non-Linear Component of Block
Cipher by means of Chaotic Dynamical System
and Symmetric Group
The fascinating attributes of chaotic and algebraic systems are found useful in the
complex field of data security. The simplest chaotic dynamical system is the double pendulum.
Here in this chapter, two double pendulums are used to enhance the chaotic behaviour of a
dynamical system. This system is sensitive to initial conditions and bears complex and chaotic
trajectory. Moreover, being multi-dimensional system it endures grander solution space for the
generation of large number of S-boxes. Furthermore, a permutation comprising on only two
cycles of symmetric group of order 256 is applied to generate integer values for the generation
of desired substitution box. The algebraic analysis of suggested S-box emphasis on its
application, thereafter, an image is encrypted with the help of this S-box, whose statistical
analysis validates its efficacy.
3.1. Contextual Review
The exploration of chaotic systems started some 200 years ago. A system whose current
state cannot be determined by initial conditions is known as chaotic system. The current state
of the system is the consequence of the past initial conditions, medium of communication, the
noise and external circumstances beyond the control of the observer. Hence randomness,
ergodicity and sensitivity to initial conditions are ultimate topographies of chaotic system.
39
These features motivates cryptographers to use such system for secure communications of
media using cryptographic algorithms.
Secure communication using wireless channels is mandatory since cryptanalysts are
always in line to extract the vital information. Thus, use of cryptography is only way to tackle
such situations. The main aim of cryptographic algorithms is to create ambiguity in the
enciphered information which is achieved using substitution boxes. These are only nonlinear
components of block ciphers generating pandemonium in cryptosystems. Many articles are
available in literature to construct such non-linear components utilizing different algebraic and
chaotic maps, some of them are listed here. [8], [30] and [31], but the chaotic dynamical
systems are utilized very often in the field of cryptography.
The motivation behind the utilization of chaotic dynamical systems like double
pendulum in the design of cryptosystems is due to the fact of the unpredictability and complex
behaviour of the system. These systems are governed by the differential equations. A physical
system is modelled initially by finding derivatives of the function. These systems are key
sensitive i.e. for a different set of initial conditions and parameters, a totally dissimilar chaotic
trajectory is obtained. Moreover, with the involvement of numerous equations and conditions,
chaotic dynamical systems are having enriched key space as compared to one dimensional
systems.
3.1. Double Pendulum
The simplest chaotic dynamical system is double pendulum. Whenever initial angels are
slightly changed, the bifurcation pattern of this system changes exponentially. Being sensitive
to initial conditions, the chaotic dynamical system is found prolific in generating confusion and
diffusion in the cryptosystem. In this article, two double pendulums having same inclinations
initially are used to generate integer values to design the non-linear components of block
40
cipher. The mathematical formulation of a double pendulum shown in Fig. (1) is explained as
follow
Fig. 1. Double pendulum attached end to end
๐ฅ1 = ๐1 sin๐1 (3.1)
๐ฅ2 = ๐1 sin๐1 + ๐2 sin๐2 (3.2)
๐ฆ1 = โ๐1 cos๐1 (3.3)
๐ฆ2 = โ๐1 cos๐1 โ ๐2 cos ๐2 (3.4)
Where ๐ฅ1and ๐ฅ2 are horizontal components and ๐ฆ1 and ๐ฆ2 are vertical components of masses
๐1 and ๐2 respectively. Now the potential energy ๐ for case of double pendulum is given as
๐ = โ๐1๐๐1 cos๐1 โ๐2๐(๐1 cos ๐1 + ๐2 cos๐2) (3.5)
And kinetic energy ๐พ is obtained by finding derivatives of Eq. (1)-(4), we get
๐พ =1
2๐1(๐1ฬ
2๐12) +
1
2๐2(๐1ฬ
2๐12 + ๐2ฬ
2๐22 + 2๐1ฬ๐1๏ฟฝฬ๏ฟฝ2๐2 cos(๐1 โ ๐2)) (3.6)
The Langrangian (L) of a system is defined as the difference of kinetic energy and potential
energy, which, for the case of a double pendulum is
๐ฟ =1
2(๐1 +๐2)๐ฟ1
2๐1ฬ2 +
1
2๐2๐ฟ2
2๐2ฬ2 +๐2๐ฟ1๐ฟ2๐1ฬ๐2ฬ cos(๐1 + ๐2)
+ (๐1 +๐2)๐๐ฟ1 cos๐1 +๐2๐๐ฟ2 cos ๐2 (3.7)
41
Then,
๐๐ฟ
๐๐1= โ๐ฟ1๐(๐1 +๐2) sin๐1 โ๐2๐ฟ1๐ฟ2 ๐1ฬ๐2ฬ sin(๐1 โ ๐2) (3.8)
๐๐ฟ
๐๐1ฬ= (๐1 +๐2) ๐ฟ1
2๐1ฬ +๐2๐ฟ1๐ฟ2 ๐2ฬ cos(๐1 โ ๐2) (3.9)
๐
๐๐ก(๐๐ฟ
๐๐1ฬ) = (๐1 +๐2) ๐ฟ1
2๐1ฬ +๐2๐ฟ1๐ฟ2 ๐2ฬ cos(๐1 โ ๐2) โ ๐2๐ฟ1๐ฟ2 ๐2ฬ sin(๐1 โ ๐2)(๐1ฬ โ
๐2ฬ) (3.10)
Since Langrangian of a system satisfies the Euler-Langrange differential equation
๐
๐๐ก(๐๐ฟ
๐๐1ฬ) โ
๐๐ฟ
๐๐1= 0 (3.11)
Substituting Eqs. (3.9) and (3.10) in above equation we get
(๐1 +๐2)๐ฟ12๐1ฬ +๐2๐ฟ1๐ฟ2๐2ฬ cos(๐1 โ ๐2) โ ๐2๐ฟ1๐ฟ2 ๐2ฬ
2 sin(๐1 โ ๐2) +
๐๐ฟ1(๐1 +๐2) sin ๐1 = 0 (3.12)
Extracting ๐1ฬ from the above equ, we get:
๐1ฬ =โ๐2๐ฟ2๐2ฬ cos(๐1โ๐2)โ๐2๐ฟ2๐2ฬ
2 sin(๐1โ๐2)โ๐(๐1+๐2) sin๐1
(๐1+๐2)๐ฟ1 (3.13)
Similarly, we can derive an equation using Euler-Langrange equation for ๐2, which is as follow
๐2ฬ =โ๐ฟ1๐1ฬ cos(๐1โ๐2)โ๐ฟ1๐1ฬ
2 sin(๐1โ๐2)โ๐ sin๐2
๐ฟ2 (3.14)
Solving above two equations simultaneously to derive the following differential equations
๐1ฬ =
โ๐2๐ฟ1๐1ฬ2sin (๐1โ๐2) cos(๐1โ๐2)โ๐2๐ฟ2๐2ฬ
2 sin(๐1โ๐2)+๐2๐๐ ๐๐(๐2)cos (๐1โ๐2)โ๐(๐1+๐2) sin๐1
(๐1+๐2)๐ฟ1โ๐2๐ฟ1cos2(๐1โ๐2) (3.15)
๐2ฬ =
๐2๐ฟ2๐2ฬ2sin (๐1โ๐2) cos(๐1โ๐2)+๐ฟ1๐1ฬ
2 sin(๐1โ๐2)(๐1+๐2)+๐๐ ๐๐(๐1)cos(๐1โ๐2)(๐1+๐2)โ๐(๐1+๐2) sin๐1(๐1+๐2)๐ฟ2โ๐2๐ฟ2cos
2(๐1โ๐2) (3.16)
Now replacing ๐1, ๐2, ๐1ฬ ๐๐๐ ๐2ฬ by ๐1, ๐2, ๐3, ๐๐๐ ๐4 respectively. Differentiation of these
yields the following four first order differential equations after substituting ๐1ฬ and ๐2ฬ:
๐1ฬ = ๐1ฬ
42
๐2ฬ = ๐2ฬ
๐3ฬ =
โ๐2๐ฟ1๐32sin (๐1โ๐2) cos(๐1โ๐2)โ๐2๐ฟ2๐4
2 sin(๐1โ๐2)+๐2๐๐ ๐๐(๐2)cos (๐1โ๐2)โ๐(๐1+๐2) sin๐1
(๐1+๐2)๐ฟ1โ๐2๐ฟ1cos2(๐1โ๐2) (3.17)
๐4ฬ =
๐2๐ฟ2๐42sin (๐1โ๐2) cos(๐1โ๐2)+๐ฟ1๐4
2 sin(๐1โ๐2)(๐1+๐2)+๐๐ ๐๐(๐1)cos (๐1โ๐2)(๐1+๐2)โ๐(๐1+๐2) sin๐2(๐1+๐2)๐ฟ2โ๐2๐ฟ2cos
2(๐1โ๐2) (3.18)
Solving the above four first order differential equations for two double pendulums in MATLAB
for 50 seconds. The graph given in Fig. 2 depicts the chaotic nature of this dynamical system.
The trajectories of two double pendulums are represented by colours in figure i.e. blue and red.
Fig. 2. Bifurcation diagram for double pendulum
The initial inclinations of two double pendulums along with initial conditions of differential
equations are responsible to determine the chaotic trajectory of dynamical system. A slight
change in their values generate a different bifurcation pattern. In other words, the solution space
is sensitive to initial keys. This concept is very useful in cryptography for the generation of S-
boxes. The robustness of the scheme based on such systems increases exponentially. For the
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 -2
-1.5
-1
-0.5
0
0.5
1
1.5
2 t = 50 s
43
case of two double pendulums, slight variation in initial parameters generated the following
different bifurcation pattern. It implies that with these values one can generate a totally different
substitution box. Hence the suggested method is key sensitive.
Fig. 3 Chaotic bifurcation pattern of system for different initial conditions
The dominance of chaotic dynamical systems over low dimensional discrete chaotic
systems is due to the fact that they have larger and complex solution space. Their larger key
space and key sensitivity are also contributing in their supremacy. Moreover, chaotic range of
continuous chaotic systems is bigger than discrete systems. Additionally, with the invention of
modern computing devices, the chance of resistance attacks like brute force etc. are minimum
for chaotic dynamical systems as compared to 1D and 2D systems.
3.2. Construction of S-Box
The confidentiality in any cryptosystem is increased utilizing substitution boxes. We suggest a
new scheme for the design of S-box based on chaotic dynamical system. The simplest chaotic
dynamical system is double pendulum. Two double pendulums making an extreme chaotic
trajectory are used to construct S-box in this scheme. The result analysis of nonlinearity, strict
avalanche criterion (SAC), bit independence criterion (BIC) and differential and linear
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2-2
-1.5
-1
-0.5
0
0.5
1
1.5
2t = 50.000 s
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2-2
-1.5
-1
-0.5
0
0.5
1
1.5
2t = 50.000 s
44
approximation probabilities validates the proficiency of the suggested S-box. Substitution box
construction involves the following steps.
Initially, from the solution space of two double pendulums a chaotic sequence ๐(1 ร
256) of integers is generated using MATLAB.
Find the sequence ๐(1 ร 256) as follow.
๐(๐) = ๐(256) โ ๐(๐) 1 โค ๐ โค 256 (3.19)
Arranging ๐(๐) in ascending order to obtain ๐(๐).
After that each element of ๐(๐) is replaced by its order in ๐(๐) to obtain ๐(๐).
A new sequence ๐โฒ(๐) is obtained by the following relation
๐โฒ(๐) = ๐(1) โ ๐(๐) 1 โค ๐ โค 256 (3.20)
Permuting the position of ๐โฒ(๐) with random permutation generated by MATLAB to
obtain the 16 ร 16 matrix ๐16 given in table 1.
Table 1: 16 ร 16 matrix
255 5 24 135 110 181 9 19 81 62 219 75 226 225 170 248
222 100 99 74 137 72 88 0 220 147 111 71 102 235 143 41
202 53 189 179 186 22 119 30 39 131 54 136 109 185 205 94
127 252 161 188 155 211 158 97 83 153 66 16 247 243 48 232
52 214 204 80 157 249 251 3 13 47 165 126 106 92 167 49
35 37 82 25 43 50 171 238 10 28 57 17 166 139 193 65
124 18 61 159 227 70 209 163 234 95 69 229 142 183 107 236
7 239 34 217 160 101 216 4 141 241 156 96 196 162 20 199
8 246 103 122 200 38 42 134 146 40 223 145 154 187 86 11
29 197 244 64 172 150 76 105 27 45 85 26 206 210 168 213
180 133 228 192 174 112 182 63 117 175 36 2 44 240 60 78
45
128 250 224 203 151 176 208 67 89 212 121 125 164 14 253 93
245 77 221 237 98 177 195 130 73 123 152 198 169 231 58 1
132 184 254 215 6 114 108 173 12 190 46 55 242 87 31 59
56 140 201 194 33 144 115 90 191 218 23 138 118 104 178 68
32 79 113 116 230 120 148 15 129 51 207 21 91 84 233 149
In the last step, a permutation ๐ from symmetric group ๐256, containing only two cycles,
is utilized for permuting the entries of ๐16. This process leads us to obtain the desired
S-box given in table 2.
๐ = (0 125 171 106 242 81 164 217 138 244 187 201 13 224 253 226 41 50 216 129
20 194 205 60 99 37 67 150 240 114 246 118 198 113 97 57 69 119 191 117 18 51 1 64
94 241 115 210 12 59 124 111 131 74 228 248 123 31 104 245 25 102 86 10 32 197 44
128 48 239 151 7 130 89 196 168 229 8 172 93 61 126 52 178 160 193 180 73 146 225
66 170 212 163 71 213 127 14 211 254 42 70 38 159 219 87 153 101 207 9 157 62 108
195 83 47 223 218 182 252 133 92 88 165 96 147 58 17 177 214 39 233 121 43 166 189
215 179 137 134 152 135 149 162 200 103) (2 155 255 91 227 26 95 145 183 65 192 105
110 49 85 72 243 45 63 174 167 132 29 176 84 22 231 53 75 156 190 33 21 79 112 158
140 36 15 186 181 208 238 30 78 5 175 230 169 6 54 209 202 40 204 188 247 251 142
55 221 232 107 235 35 143 56 34 19 220 4 236 122 109 3 11 234 68 139 46 24 120 250
27 206 185 154 249 80 100 77 28 16 90 237 136 161 76 199 148 184 173 23 82 116 144
222 203 98 141)
Table 2: Designed S-box
163 188 187 183 242 167 175 105 144 1 171 135 254 123 199 186
102 66 101 189 246 53 43 240 205 120 194 207 92 200 178 96
46
219 253 83 138 154 159 251 108 152 201 233 241 177 116 145 25
8 107 131 99 20 90 9 86 11 18 64 226 231 139 210 185
5 67 140 22 23 153 54 192 50 151 122 211 133 227 143 72
51 113 0 237 128 252 209 55 166 103 222 149 38 44 52 111
112 239 198 247 35 45 147 73 41 245 2 191 48 156 95 196
49 195 32 79 82 93 148 249 220 218 118 129 16 255 243 114
109 12 19 236 63 91 40 27 104 203 190 157 168 61 21 179
160 65 47 124 130 134 80 68 42 238 212 24 126 62 7 119
224 146 150 6 37 10 136 60 98 115 204 162 81 89 232 181
235 100 155 173 77 121 46 197 172 58 248 230 169 182 206 4
214 180 137 142 28 202 216 106 228 125 184 31 39 221 71 70
176 158 217 170 36 3 250 14 161 174 59 213 74 97 29 94
225 244 84 17 165 117 78 88 87 30 75 229 110 57 132 127
76 193 69 13 223 234 34 164 33 85 141 15 208 56 215 26
3.3. Analysis of S-box
Nonlinearity
It is the most imperative property of a cryptosystem. Principally, the nonlinearity of an
outstanding cryptographic system is higher. It measures the confrontation of a system being
expressed as a set of linear equations and hence confirms resistance against linear
cryptanalysis. Using theory of Boolean functions, for a Boolean function ๐ข, the nonlinearity is
defined as follow:
๐๐ฟ๐ข = ๐(๐ข, ๐๐) = min๐ (๐ข, ๐ฟ); ๐ฟ โ ๐๐ (3.21)
where ๐๐ is the collection of affine Boolean transformations.
Table 3: Comparison table of nonlinearity values of various S-boxes
47
S-box Suggested Ref. [8] Ref. [35] Ref. [34] Ref. [32] Ref. [36] Ref. [2]
Average 111.5 106.75 103.3 103 104.7 105.25 112
Minimum 110 106 99 100 102 102 112
Maximum 112 108 106 106 108 108 112
Bit Independence Criterion
The statistical property of output bit independent criterion (BIC) for an S-box given by Webster
and Tavares [23] is delineated as, for a certain collection of avalanche vectors, altogether the
avalanche variables should be pairwise autonomous. This principle gives the impression to
highlight the proficiency of the confusion function.
Table 4: Comparison table of BIC-SAC values for various S-boxes
S-box Suggested Ref. [8] Ref. [35] Ref. [34] Ref. [32] Ref. [36] Ref. [2]
BIC-SAC 0.5053 0.4989 0.4987 0.5050 0.5058 0.4956 0.504
BIC
nonlinearity
111.357 106.6 103.3 103.1 104.1 103.8 112
Strict Avalanche Criterion
The Strict avalanche effect (SAC) introduced by Webster and Tavares in 1985, is basically the
generalization of completeness and avalanche effect. It is defined as all the output bits differ
by a probability of a half whenever single input bit is complemented. Mathematically, for
โ: ๐บ๐น(2)๐ โ ๐บ๐น(2)๐
๐๐๐๐(โ(๐ฅ๐))๐ โ ๐๐๐๐(๐(๐ฅ))๐=1
2 โ ๐ โ [1, ๐] ๐๐๐ ๐ โ [1,๐] (3.22)
Table 5. Table of comparison for SAC outcomes
S-box Suggested Ref. [8] Ref. [35] Ref. [34] Ref. [32] Ref. [36] Ref. [2]
48
Minimum 0.4375 0.4219 0.4140 0.4218 0.3906 0.4297 0..453
Average 0.5053 0.4939 0.499 0.500 0.506 0.496 0.504
Maximum 0.5781 0.5625 0.6015 0.6093 0.5937 0.5313 0.562
Linear and Differential Approximation Probabilities
The resistance of an S-box against linear and differential attacks is accessed in these analyses.
The linear approximation probability (LAP) measures the largest amount of the disproportion
of an event. Let ๐ represents collection of all probable inputs having total components 2๐. If
input and output are ๐๐ ๐๐๐ ๐๐ respectively, then LAP values compared in Table 6 are found
using the equation:
๐ฟ๐ด๐ = max๐๐,๐๐โ 0
|#{๐ข/๐ข. ๐ข = ๐(๐ข). ๐๐ }
2๐โ1
2| (3.23)
Table 6: Comparison table of LAP values for different S-boxes
S-box Suggested Ref. [8] Ref. [35] Ref. [34] Ref. [36] Ref. [32] Ref. [2]
Max. LP 0.0703 0.1250 0.1328 0.1289 0.1562 0.1250 0.062
Max.
Value
146 160 164 162 168 160 144
The differential approximation probability (DP) also known as differential homogeneity is an
ultimate trait of a substitution box. For a robust S-box the outcomes of DP must be close to
zero. The comparison of DP values of the proposed S-box with various S-boxes is given in
Tale 7.
Table 7:. Comparison table of DP values for different S-boxes
S-box Suggested Ref. [8] Ref. [35] Ref. [34] Ref. [36] Ref. [32]
Max. DP 0.01563 0.0625 0.03906 0.05469 0.03906 0.04688
49
3.4 Majority Logic Criterion
Majority logic criterion (MLC) is used to gauge the efficacy of an S-box as a replacement of
nonlinear block ciphers component. It has to produce promising outputs of different analyses
used in MLC to qualify for this position in block ciphers. It includes the comprehensive debate
on entropy analysis, contrast, homogeneity, correlation and energy analysis [25]. Some very
brief details of these analysis are given hereafter.
Information Entropy Analysis
The statistical amount of disorder in the cipher image gives the information entropy. If the total
number of pixels in image are ๐, then measure of the information entropy of an enciphered
image is:
E(U) =โ๐(๐ข๐) log2 ๐(๐ข๐)
๐
๐=1
(3.24)
Where, ๐(๐ข๐) is the probability of the pixel ๐ข๐ . The hypothetical value of entropy for a grey
scale image is 8, when the chance of pixel occurrence is same. Hence, the information entropy
for the suggested scheme must be nearest to 8, to validate its efficacy. Table 11 gives the
comparison of entropy analysis of the proposed scheme with [27], [8] and [36].
Correlation Analysis
The adjoining pixels (horizontally, vertically and diagonally) of the host image are highly
correlated. A secure and robust encryption procedure make these adjacent pixels unrelated, i.e.
the correlation of adjacent pixels approaches to zero of an enciphered image. The following
formula is used to calculate the correlation between adjacent pixels ๐ ๐๐๐ ๐ of the grey level
image
50
ฮ๐๐ =โ ({๐๐ โ ๏ฟฝฬ ๏ฟฝ}{๐ ๐ โ ๏ฟฝฬ ๏ฟฝ})๐๐=1
โโ (๐๐ โ ๏ฟฝฬ ๏ฟฝ)2๐๐=1 โโ (๐ ๐ โ ๏ฟฝฬ ๏ฟฝ)2
๐๐=1
(3.25)
The outcomes generated by above relation are shown in Table 8. The correlation coefficient of
the plaintext image and enciphered image is nearly equal to one and zero respectively as
desired.
Table 8. MLC comparison table of proposed S-box for the Cameraman Image.
Images Entropy Correlation Energy Homogeneity Contrast
Original 7.1025 0.9292 0.1679 0.8964 0.4785
Proposed S-box 7.9845 0.0023 0.0157 0.3952 10.2584
Ref. [27] 7.9591 -0.0441 0.0202 0.4151 8.2314
Ref. [8] 7.9812 -0.0045 0.0177 0.4091 8.3154
Ref. [36] 7.9431 0.0155 0.0219 0.4248 8.2113
Ref. [2] 7.9561 0.0554 0.0202 0.4662 8.3124
Contrast, Homogeneity and Energy Analyses
An appropriate amount of brightness is present in the host image, which vanishes in the
enciphered image. This loss is measured by the contrast analysis. The secure encryption yields
the higher values of contrast for the encrypted image. Besides the behaviour and characteristics
of an encrypted image is gauged using the analyses of energy and homogeneity.
51
Fig. 4. Host Image and the Encrypted Cameraman image.
Fig. 5. Histogram of plaintext Image and the Encrypted image.
To increase the vagueness of a cryptosystem, use of chaotic maps in construction of substitution
box is very common now a days. In this paper, the simplest chaotic dynamical system i.e.
double pendulum is used for the first time to generate integer values along with the application
of symmetric group in construction of non-linear components of block cipher. The
amalgamation of these two yields confusion and diffusion in the suggested cryptosystem. For
practical application, an image is encrypted afterwards with this cryptosystem. The standard
algebraic and statistical analyses available in literature validate the efficacy of the proposed
system for the safe communication of data. Hence, designed chaotic S-box generated by means
of chaotic dynamical system and symmetric group is the main hurdle in the path of
cryptanalysts.
52
Chapter 4
Applications of Nonlinear Coupled Differential
Equations in Multimedia Security
This chapter focuses on multimedia applications like image encryption and watermarking
scheme using systems of nonlinear coupled differential equations. Rabinovich-Fabrikant (RF)
system of differential equations contains nonlinear term of cubic order which makes the system
more complex. The dynamics of RF system in consequence of this becomes chaotic. Moreover,
different chaotic solutions can be achieved by slightly varying parameters and conditions. It is
a three dimensional system hence generates three chaotic sequences of pseudorandom
numbers. These three layers are to be used for colour image encryption and watermarking
schemes separately. Three main operations are involved including permutation, substitution
and XOR. The processed images are found robust and strong when they were testified by
different algebraic and statistical analyses. Moreover, NIST test for randomness was also found
successful for encrypted images.
4.1 Introduction
The significance of internet and computer in speedy communication is well well-known datum
of this modern day world. The general civic from various sections of society are believing such
insecure mode of communications reluctantly. With the invention of soft computing devices,
stress-free and speedy communications is becoming one of the basic needs of the populaces.
This ease in communication increases the security threat to the valuable information being
53
transmitted. Such safety concerns include the illegitimate access to the data, alteration of the
data and unauthorised ownership claims by the assaulters [15].
Cryptography gives solution to this problem by developing schemes for secure communication.
Chaos and cryptography have some traits in common like sensitivity towards conditions,
random behaviour and ergodicity. Both of these are closely associated after the theory of
Shannon [1]. Simple mathematical chaotic maps like logistic, tent and piecewise linear maps
are very common in cryptography [8], [30] and [31]. These are easy to use and implement in
the design of cryptosystems but can be attacked as well due to their simple structure. To
overcome this hurdle chaotic and complex systems like nonlinear differential equations can be
used.
Data communicated by end users includes text memos, voice messages and images. Images
and pictures being different from text are the major portion of communication among
populaces. Pixels are the building blocks of images, they are arranged in a definite pattern to
form a proper texture and shape of an image. Therefore, their security demands extra ordinary
procedure of encryption and watermarking. Literatures reveals many articles on image
encryption [39], [40], [41] and [42].
The scientific layout of chaotic dynamical frameworks some of the time brings about the type
of ordinary differential equations (DE). For a chaotic dynamical structure, the adjustment in its
capacities is assessed by standard or incomplete subordinates to plan an administering issue.
This speculation is utilized to figure out the conduct of an obscure ability sooner rather than
later. A portion of the nonlinear DE produce nonlinear behaviour and such frameworks are
discovered sensitive towards introductory information. These qualities are suitable for the
structure of secure correspondence arrangement in cryptography [21], [22], [27], [28] and [29].
Rabinovich-Fabrikant (RF) arrangement of DE is among one of those that can be utilized
rightly for the design of secure cryptosystem.
54
4.2 RabinovichโFabrikant Equations
RF generates a chaotic as well as complex bifurcation pattern [18]. It is composed of three
coupled ordinary differential equations containing two fixed parameters and three variables as
explained in chapter 2, section 2.2. Here, we will use this system for application in image
encryption and water marking scheme. The dominance of the suggested system is due to the
fact of generating three pseudorandom numbers sequence from a single mathematical system.
Moreover, the chaotic behaviour of the system is enriched due to the involvement of cubic
order nonlinearity. Additionally, it has the ability to generate different sequences by slightly
varying parameters and initial conditions.
It is very often observed that combined differential equations have chaotic behaviour. In the
case of RF-system, the nonlinearity of third order results as chaotic and complex dynamics of
this system. This in result creates the difficulty in investigation of analyses like hetroclinic or
homoclinic orbits and existence of invariant sets etc. Moreover, utilization of various
software/techniques operating on dissimilar step size produces different chaotic pattern [33].
-4
-2
0
2
-4
-2
0
2
4
0
0.5
1
1.5
2
RF system sigma=0.100000000, beta=0.140000000
-2
-1.5
-1
-0.5
0
-1
0
1
2
3
0
0.5
1
1.5
RF system sigma=0.87654, beta=1.11234
55
Figure 1. Different chaotic behaviour of RF system when sigma and ๐ฝ are changed
4.2.1. The Literature Review
The significance of pictures in correspondence can't be denied by any means. A gigantic
measure of information as imaginings is circled by people in general just as the associations.
A portion of the touchy and significant pictures require a safe method of transmission.
Numerous encryption calculations are accessible in writing to accomplish the ideal mystery. A
not very many are recorded in the following paragraph.
Like Turan et.al in [39] in the year 2018 proposed an encryption scheme using total diffusion.
For the creation of stream of pseudorandom numbers, they utilised logistic map to make it
applicable in encryption scheme. Wang et.al amalgamated coupled map lattice (CML) and
DNA sequences in planning image encryption system in [40]. Primarily, plaintext image is
enciphered using DNA code to generate a DNA matrix. This obtained matrix is then scrambled.
Lastly, jumbled matrix is diffused via CML and DNA sequences to produce an enciphered
image.
Ahmed et.al suggested a cryptographic technique based on 4D hyper chaotic framework in
[43]. An encryption scheme dependent on chaotic map and S-box was developed in [41]. In
[42], authors utilized 3D nonlinear maps for the plan of encryption procedure. Waseem et.al
utilized a thought of quantum turning and rotation to accomplish the protection of pictures in
correspondence [44].
-2
0
2
4
-4
-2
0
2
4
0
0.5
1
1.5
RF system sigma=0.1, beta=0.2715
-4
-2
0
2
4
-4
-2
0
2
4
0
1
2
3
4
5
RF system sigma=-1, beta=-0.1
-2
-1
0
1
2
-5
0
5
-0.5
0
0.5
1
1.5
RF system sigma=0.1, beta=0.98
56
Zahmoul et.al produced random chaotic maps using beta functions for the generation
of chaotic sequences in [45]. This cryptosystem was based on transformation, diffusion and
switching of digits by new generated sequences. In [46], authors used two mathematical
structures i.e. linear fractional transform and lifting wavelet transform for partial image
encryption scheme. Two dimensional chaotic cat map was used by Safwan et.al along with bit
level permutation for the encryption scheme in [47]. DNA and chaos based system was used
by Mondal et.al in [48] to propose lightweight but yet effective and secure cryptosystem. Three
dimensional cat map along with Chen chaotic system were used by authors in [49] to suggest
three dimensional bit matrix permutation for encryption of images. Chai et.al proposed the
same concept of DNA sequences operations along with 2D logistic map for image privacy
preserving scheme in [50]. Cavusoglu et.al suggested secure image encryption outline based
on only chaotic S-box in [51]. In [52] Li X et.al presented a colour image encryption procedure
based on complex chaotic system and DNA. Moysis et.al applied the idea of 2D chaotic maps
for the encryption algorithm [53]. There are many similar encryption techniques available in
literature that are developed on chaotic systems like [54], [55] and [56].
4.3 Proposed Image Encryption Scheme
Ongoing advancements in delicate figuring gadgets changed the all-out conveying conduct of
people in general. Various important and sensitive information is communicated in terms of
images sometimes. Pictures containing official and touchy data identified with military,
medicinal and bookkeeping segments and so on, eventually requires secure correspondence.
This objective can be accomplished by creating cryptosystem technique that can guarantee
secure correspondence between any two customers.
57
The suggested encryption system involves four stages, i.e. production of chaotic sequences,
permutation phase, bitwise pixels Xoring phase and the substitution phase. This algorithm is
explained stepwise in the following manners:
E1: Firstly, RF system is utilized to generate three different chaotic structures ๐, ๐ ๐๐๐ ๐ of
pseudorandom numbers and then permuted to improve their randomness.
E2: In the second step, a colour image of size 256 ร 256 is separated into its three stratums
i.e. blue, green and red.
E3: In this step, each layer of the coloured image is permuted using MATLAB.
E4: In the fourth step, the matrices of red, green and blue channels from step E3 and the
sequences ๐, ๐ ๐๐๐ ๐ from step E1 are xored bitwise.
E5: Before the final step, i.e. in the substitution process, the pixel entries of green, blue and red
matrices are substituted using the sequences ๐, ๐ ๐๐๐ ๐ respectively.
E6: Lastly, the three matrices of red, green and blue layers of substituted matrices are combined
to generate a coloured encrypted image as depicted in Figure 2.
Figure 2. Plaintext Lena image along with Enciphered Image of size 256 ร 256.
58
4.4 Security Standards to gauge the strength
4.4.1. Key Space
Big key space is one of the reason found responsible for creating resistance against the different
brute force viciousness. It is naturally difficult to create/design the accurate keys for the
decryption of an image from a large key space. For a key of length ๐, total key space for image
encryption would be of length 2๐. Thus key space upsurges very sharply whenever a small
integral increment is given to ๐. The key space in this algorithm depends upon the selection of
beta, sigma and three initial conditions ๐ฅ0, ๐ฆ0, ๐ง0. The computational accuracy presented by
[57] in IEEE floating-point standard is 1015, then total possibilities of key in this case would
be 1015ร5 โ 2230
59
Chaotic sequence for RF system
Sequence YSequence X Sequence Z
Original Colour Image
Red Layer Green Layer Blue Layer
Bitwise XOR
Permutation
Phase
Substitution
Bitwise XOR
Permutation
Phase
Substitution
Bitwise XOR
Permutation
Phase
Substitution
Encrypted Image
Figure 3. Flow Chart for the Suggested Image Encryption Scheme
4.4.2. Key Sensitivity Analysis
A chaotic system is imperatively sensitive on initial conditions and parameters. If either
initial conditions or constraint is given an increment, a chaotic system will yield totally
different phase portrait which also supports in image encryption as even a very minute wrong
guess of keys will hinder decryption. The mathematical scheme known as Runge-Kutta of order
4 is used to produce chaotic sequences from coupled differential equation in the proposed
algorithm. Which is very sensitive for keys and confirms a wrong initial guess or step size used
in numerical scheme will generate a completely different picture. Therefore, the proposed
cryptosystem is extremely sensitive as for as the case of keys are concerned.
60
4.4.3. Complexity Analysis
The two important parameters studied in this analysis are space and time taken by the algorithm
on a computing device. The computing machines nowadays, are having large enough
memories, therefore the issue of space is minimised. So our problem in terms of complexity is
concerned only with execution time. If we run the suggested scheme on a laptop with enhanced
RAM and CPU, its average execution time will be much higher as compared to slow computing
devices. We worked on a fifth generation laptop having 4GB RAM and 3.00GHz CPU and
found execution time as 218 kb/sec, which is much higher than [36] and [58] whose average
finishing time as observed is 166 kb/sec and 125 kb/sec respectively.
(c). Original Red Lena Image (d). Original Green Lena Image (e). Original Blue Lena Image
(h). Encrypted Red Image (g). Encrypted Green Image (h). Encrypted Blue Image
Figure 4. Original and Encrypted Red, Green and Blue Lena Images (a)-(h).
61
4.4.4. Information Entropy
The mathematical formula for information entropy (IE) which is also known for the evaluation
of uncertainty in a random variable is defined by
๐ผ๐น๐ฟ = โ๐(๐ฟ = ๐)
๐นโ1
๐=0
๐๐๐21
๐(๐ฟ = ๐) (4.1)
Here P denotes pixels percentage and F represents gray level value. The ideal value of IE for a
gray-scale image is 8. Any result closer to this value implies the robustness of the scheme. IE
values of the proposed scheme in contrast with some Ref. [36], [58], [55] are tabulated in table
1.
Table 1. Entropy outcomes of Baboon and Lena images.
Image Layer Suggested Ref [7] Ref [36] Ref [55] Ref [58]
Baboon Coloured
Image
R 7.9977 7.9973 7.9987 7.8124 7.9981
G 7.9985 7.9969 7.9989 7.8937 7.9974
B 7.9989 7.9985 7.9990 7.8561 7.9977
Lena coloured
Image
R 7.9976 7.9972 7.9992 7.8834 7.9986
G 7.9982 7.9977 7.9991 7.8756 7.9972
B 7.9984 7.9987 7.9988 7.8907 7.9968
4.4.5. Correlation Analysis
The main purpose of image encryption techniques is to distort the relationship between the
adjoining pixels of an original image as these are extremely correlated with each other in all
three directions to generate a distorted and unrecognizable enciphered image. The correlation
between adjoining pixels must be approaching towards zero in a secure encryption. For a two
neighbouring pixels ๐ฅ and ๐ฆ the formula for correlation is as follow
62
๐ถ๐ฅ๐ฆ =[๐ธ(๐ฆ โ ๐๐ฆ)(๐ฅ โ ๐๐ฅ)]
๐๐ฅ๐๐ฆ (4.2)
Whereas ๐ธ[. ], ๐, ๐๐ฅ๐๐๐ ๐๐ฆ are estimated value, mean and standard deviation in ๐ฅ and ๐ฆ
directions respectively.
Correlation diagonally Correlation horizontally Correlation vertically
Figure 5. Correlation for diagonal, horizontal and vertical directions for original image
Correlation diagonally Correlation horizontally Correlation vertically
Figure 6. Correlation for all three directions for an encrypted image.
Table 2. Correlation of plaintext and enciphered text images in all three directions.
Image Layer Diagonal Horizontal Vertical
Original Baboon
Image
R 0.9006 0.9414 0.9228
G 0.8203 0.9033 0.8842
B 0.9059 0.9438 0.9331
R โ0.0082 โ0.0023 0.0090
63
Encrypted Baboon
Image by proposed
Scheme
G 0.0065 โ0.0068 โ0.0091
B 6.124 ร 10โ4 โ0.0052 โ0.0025
Original Lena Image
R 0.9519 0.9753 0.9838
G 0.9421 0.9735 0.9820
B 0.9272 0.9541 0.9698
Encrypted Lena
Image by Proposed
Scheme
R 0.0066 โ0.0033 โ0.0044
G โ0.0084 0.0089 3.375 ร 10โ4
B โ0.0080 0.0003 0.0007
Ref [55] 0.0045 โ0.0012 โ0.0041
Encrypted Baboon
Image by Ref [36]
R โ0.0016 โ0.0072 โ0.0201
G โ0.0175 โ0.0260 โ0.0220
B โ0.0066 โ0.0099 โ0.0034
Ref [59] 4.001 ร 10โ4 0.0038 0.0023
4.4.6. Histogram Analysis
Histogram analysis is used to calculate similarities between the pixels for observing changes
in the image due to substitution process. Histogram of an encrypted image is usually evenly
distributed as compared to original image. This means the more the uniformity in the histogram
of an encrypted image the more protected is the cryptosystem.
64
Figure 7. Histogram of original and enciphered Lena image
4.4.7. Sensitivity Analysis
The sensitivity of an encryption scheme is satisfied whenever an observer observes huge
alterations and randomness in output file for a minute increment input. In view of images, it
implies for an increment in single pixel of an enciphered image must result maximum pixels
alterations. This confirms the resistance and strength of the scheme against differential
analysis. There are major tests that are used to gauge this strength named as number of pixels
change rate (NPCR) [37] and unified average changing intensity (UACI) [15].
The evaluation of response happened in an encrypted image due to the single pixel increment
is witnessed in UACI. For two such enciphered images ๐ธ1(๐, ๐) and ๐ธ2(๐, ๐) having dimensions
๐ ร๐ป is tabulated in table 3 and defined by the following formula
๐๐ด๐ถ๐ผ =1
๐ ร ๐ป|๐ธ1(๐, ๐) โ ๐ธ2(๐, ๐)
255| ร 100% (4.3)
NPCR gauges pixels change rate in comparison to the original image. For the images ๐ธ1(๐, ๐)
and ๐ธ2(๐, ๐), as defined in UACI, the formula for NPCR is
๐๐๐ถ๐ =โ ๐ถ(๐, ๐)๐,๐
๐ ร๐ปร 100% (4.4)
Where,
๐ถ(๐, ๐) = {0 ๐๐ ๐ธ1(๐, ๐) = ๐ธ2(๐, ๐)
1 ๐๐ ๐ธ1(๐, ๐) โ ๐ธ2(๐, ๐)}
65
Outcomes of NPCR are expressed for enciphered images in table 4.
Table 3. UACI results comparison table for proposed algorithm.
Image Level Suggested Average Ref [60] Ref [54] Ref [55] Ref [58]
Baboon
Coloured
Image
R 33.53 %
33.56%
32.51%
33.44%
33.54%
33.65 %
G 33.55 % 33.59 %
B 33.62 % 33.67 %
Lena
coloured
Image
R 33.60 %
33.60%
32.87%
33.76%
33.31%
41.10 %
G 33.59 % 36.58 %
B 33.63 % 32.90 %
Figure 8. Encryption of Baboon Image of size 512 ร 512
Table 4. Comparison table of NPCR values of proposed scheme with various encryption schemes.
Image Level Suggested Average Ref [60] Ref [54] Ref [55] Ref [58]
Baboon
Coloured
Image
R 99.60%
99.60%
85.92%
99.10%
91.87%
99.62 %
G 99.59% 99.63 %
B 99.62% 99.63 %
R 99.62%
99.61%
86.68%
99.60%
92.23%
99.62 %
G 99.60% 99.62 %
66
Lena
coloured
Image
B 99.63% 99.62 %
4.4.8 Randomness Test (NIST SP 800-22) for Cipher
National institute of standards and technology in their special issue presents a new and reliable
algorithm for the evaluation of randomness of sequences. It includes a list of many small tests
like block frequency, long runs of ones, spectral DFT and cumulative sums in both forward
and reverse directions. To gauge the randomness of enciphered images we have used this test.
A baboon image of size 512ร512 is initially encrypted by the proposed scheme based on RF
system as given in Fig. 3. Afterwards its random behaviour is observed using NIST [26]. The
encrypted image cleared all the tests as shown in table 5, confirming the robustness of the
proposed scheme.
Table 5: NIST test results for Encrypted RGB Baboon Image
Test P โ values for colour encryptions of
ciphered image
Resul
ts
Red Green Blue
Frequency 0.15473 0.12126 0.63526 Pass
Block frequency 0.7257 0.25553 0.57578 Pass
Rank 0.29191 0.29191 0.29191 Pass
Runs (M=10,000) 0.54801 0.60822 0.12038 Pass
Long runs of ones 0.7127 0.7127 0.7127 Pass
Overlapping templates 0.85988 0.81567 0.85988 Pass
67
No overlapping
templates
0.99981 1 0.95715 Pass
Spectral DFT 1 0.30979 0.66336 Pass
Approximate entropy 0.022869 0.66195 0.17805 Pass
Universal 0.99339 0.98605 0.99214 Pass
Serial p values 1 7.2376e-06 0.31004 0.11931 Pass
Serial p values 2 2.4517e-06 0.85774 0.36824 Pass
Cumulative sums
forward
0.18702 0.27914 0.27743 Pass
Cumulative sums reverse 0.78937 1.7241 0.92583 Pass
Random excursions X = -4 0.65048 0.86404 0.99249 Pass
X = -3 0.65189 0.52279 0.98815 Pass
X = -2 0.68678 0.1926 0.97465 Pass
X = -1 0.52847 0.15359 0.8282 Pass
X = 1 0.74765 0.54347 0.0937 Pass
X = 2 0.45619 0.5065 0.9827 Pass
X = 3 0.87117 0.7286 0.98815 Pass
X = 4 0.68706 0.087556 0.99019 Pass
Random excursions
variants
X = -5 0.2291 0.19647 0.68309 Pass
X = -4 0.2067 0.25684 0.64343 Pass
X = -3 0.16928 0.4674 0.58388 Pass
X = -2 0.35454 0.66501 0.4795 Pass
X = -1 0.6885 0.45325 0.68309 Pass
X = 1 0.10881 0.45325 0.68309 Pass
68
X = 2 0.53709 0.5164 0.09896 Pass
X = 3 0.59068 0.28818 0.017622 Pass
X = 4 0.39054 0.42187 0.030754 Pass
X = 5 0.50404 0.73888 0.13442 Pass
4.5 Watermarking Scheme
An extensive increase in soft computing devices and global networking suggested copious
overtures for the design and demonstration of the digital content. The obtainability and easy
access to digital data involving digital repositories, e-advertising, audio, video and e-libraries
etc. arose many security alarms. The existing copyright rulebooks are susceptible for the
alteration of data in the form of plagiarism as well as copyright violations are observed while
the communication of digital data on the internet. This scenario motivated many researchers to
make use of watermarking (a digital signal is hidden in the message authenticating originality
of the shared content) schemes for the authenticated transmission of digital contents like audio,
video, digital images and texts etc.
Spatial domain [61] and frequency domain [62] are the two main categories of watermarking
techniques developed yet. The former replaces the pixels of host image with the watermarked
image, inferring the greater capability to insert watermark while in the latter watermarking is
done on the coefficientsโ value of the host image yielding strong robustness against malicious
attacks. The aim of both is common i.e. to give integrity, copyright protection, authentication,
broadcast monitoring and robustness against mischievous attacks [63]. The watermarking
technique suggested based on RF-system involves the following embedding and extraction
procedures.
4.5.1 Embedding Process
The embedding process of logo image into the host image is explained in the following steps
69
Choose a host image of dimension ๐ ร๐ and separate it into three layers i.e. red green
and blue for the application of embedding process of logo image.
Generate three sequences ๐, ๐ ๐๐๐ ๐ from the solution space of RF system and permute
three layers of image using these sequences with the help of following equations
๐๐ ,๐บ,๐ต(๐, ๐) =
{
๐ผ๐ ,๐บ,๐ต(๐ โ ๐ฅ๐, ๐ โ ๐ฅ๐) ๐๐ ๐ โ ๐ฅ๐ โฅ 1; ๐ โ ๐ฅ๐ โฅ 1
๐ผ๐ ,๐บ,๐ต(๐ โ ๐ฅ๐, ๐ โ ๐ฅ๐) ๐๐ ๐ โ ๐ฅ๐ โฅ 1; ๐ โ ๐ฅ๐ โฅ 1
๐ผ๐ ,๐บ,๐ต(๐ + ๐ฅ๐ , ๐ + ๐ฅ๐) ๐๐ ๐ + ๐ฅ๐ โค ๐ ; ๐ + ๐ฅ๐ฝ โค ๐
๐ผ๐ ,๐บ,๐ต(๐ + ๐ฅ๐ โ๐, ๐ + ๐ฅ๐ โ๐) ๐๐ ๐ + ๐ฅ๐ โฅ ๐ ; ๐ + ๐ฅ๐ โฅ ๐
(4.5)
Choose a logo image of size ๐พ ร ๐พ and convert its most significant bits (MSBs) to least
significant bits (LSBs) using shift process to obtain 2๐พ ร 2๐พ LSBs.
Then select a block of size 2๐พ ร 2๐พ from permuted image and replace LSBs obtained
in the last step to the selected block.
Finally, apply inverse permutations on the processed image of previous step using the
inverse equations for permutations to obtain watermarked image. The whole scheme of
watermarking is explained in the Fig 9.
4.5.2 Extraction Process
It is an important task to ensure copyright protection is to extract the logo image appropriately.
This also ensures the productivity of scheme. The extraction process is explained as follow
Separate the watermarked image into three channels and apply permutations to each
channel through generated sequences ๐, ๐ ๐๐๐ ๐.
Choose block of size 2๐พ ร 2๐พ from permuted channel and split it into LSBs and
MSBs.
Obtain ๐พ ร ๐พ LSBs and MSBs from the above step.
Finally, apply the inverse permutations on ๐พ ร ๐พ LSBs to recover the logo image from
watermarked image.
70
Start
Blue
Permutation
System of ODEs
GreenRed Sequences
Generated by
Xยด, Y ฬand Zยด
Permute
Red
Red Sub
Block
Logo Image
Red
MSBs LSBs
LSBs
Convert MSBs to LSBs
Rearrange
the block
Red
LSB
embedding
Inverse
Permutation
Watermarked
Image
Fig. 9: Flow chart for the anticipated watermarking scheme
Fig. 10 Processed watermarked images
71
Fig. 11: Original and the extracted logo images
Fig. 12: Histograms for the Original and the watermarked images
4.6 Simulation Results and Statistical Analysis
This part of the chapter provides the detail of necessary analyses to authenticate the usefulness
of the suggested watermarking scheme. Among these are peak signal to noise ratio and mean
square error. The gray level co-occurrence matrix (GLCM) of the watermarked and original
image is examined to evaluate these. The size of watermarked image is 512 ร 512 and that of
original image is 512 ร 512.
The strength of watermarking is established by analyzing robustness, invisibility, security and
capacity.
4.6.1 Mean Squared Error (MSE)
This analysis is used to measure dissimilarity between original and the processed images. MSE
is tabulated in Table 6 and its mathematical representation is given as follow:
๐๐๐ธ =1
๐โ(๐ฅ๐ โ ๐ฅ๐
โ)2 (4.6)
72
4.6.2 Peak Signal to Noise Ratio (PSNR)
PSNR is defined as
๐๐๐๐ = 10๐๐๐10๐๐ด๐๐ผ
2
๐๐๐ธ (4.7)
Where denominator of the above equation is dissimilarity of the images i.e. MSE and
numerator is signal strength. PSNR and SSIM are used to measure similarity between Host
image and the watermarked image which is actually the evaluation of invisibility [64] and [65].
Table 6: Performance Analyses for the three layers of watermarked image
Red Green Blue
PSNR 41.4413 43.2050 42.7895
MSE 4.6660 0.1291 3.4207
NPCR 0.5767 0.587 0.05837
UACI 0.01763 0.014134 0.01484
NCC 1.002282 1.0023760 0.99921
AD average difference -0.43899 -0.29938 0.091163
SC Structural content 0.9953 0.9950236 1.001285
MD Mean diff 10 13 15
NAE Normalized absolute
error
0.002490 0.003647 0.003575
Table 7: Performance Analysis for Host and the Extracted logo QAU image
Red Green Blue
PSNR 99 99 99
MSE 1 1 1
NPCR 0 0 0
UACI 0 0 0
NCC 1 1 1
AD 0 0 0
SC 1 1 1
MD 0 0 0
NAE 0 0 0
73
4.7 Robustness Test on Image Processing Operations
The mathematical calculation of two watermarks is the similarity amongst the original and the
extracted watermark [66]. These two satisfy the criterion for robustness if they are highly
correlated. Mathematically this means that the similarity outcome must be towards higher side.
If the ๐๐กโ element of the original and extracted watermark are ๐๐ and ๐๐ then similarity is defined
as follow:
๐๐๐ =โ๐๐. ๐๐
โโ๐๐2โ๐๐2
The similarity index for secure and robust watermarking should be ideally near to 100. For the
proposed scheme it comes out to be 99.92 which is very close to the optimal value. For different
images similarity analyses are tabulated in Table 6. Several other image processing attacks are
discussed in the coming subsections.
4.7.1 Noise Attack
To measure the resistance against the noise attack, we added salt and pepper noise in the
watermarked image thereafter the logo image is extracted from the noisy image. It is evident
from Fig. 13 that the extracted logo image form the noisy watermarked image contains not
more than the inserted amount of noise. Additionally, the extracted image is undistorted apart
from the noise confirming the strength of anticipated scheme. This claim can also be visualized
form the Fig. 13 and by the results given in Table 8 for noise analysis. In parallel to this
Gaussian, Speckle, and Poisson can also be utilized for this attack.
a b
Fig. 13: Extracted logo images with (a) 2% (b) 10% noise added.
Table 8: Statistical Analysis for the Host and the watermark loge image QAU
74
Salt & Pepper PSNR MSE NPCR UACI
2% 25.5882 179.579 0.04846 0.009022
10% 18.3747697 945.37 0.2312 0.045311
4.7.2 Cropping Attack
In this attack either lesser information than the host image is offered or the distortion of
extracted image is done. Although the extraction of watermark is a difficult job whenever
various types of cropping occurred generally however this does not work in the case of
proposed watermarking scheme. All possible cuts are applied on the watermarked image and
astonishingly the extracted logo image is recognizable even in the case of different half cuts as
depicted in Fig. 14. The results of almost all attacks of image processing are tabulated in Table
9.
Fig. 14: Cropping analyses images
Table 9: Statistical Analysis for the Host and the watermark loge image QAU
75
Scaling PSNR MSE NPCR UACI
Half Left Cut 10.44862 5864.324 79.44335 20.046
Half Right Cut 10.43972 5876.355 79.7668 20.078
Half upper cut 10.36612 5976.7818 78.5400 19.6887
Half Lower Cut 10.3028 6064.5375 78.2958 20.438
Mid Cut 14.30113 2415.27 50.87 10.3383
The secure communication in the current scenario of modern-day world is a dream of
every individual, which should be protected by all means. Encryption and digital watermarking
are indeed appropriate choices for the secure communication and copyrights protection
respectively. For secure transmission of digital images, initially, an encryption scheme is
proposed in this chapter. The deployment of RF nonlinear system having complex and chaotic
behaviour due to the involvement of quadratic and cubic terms is found responsible for creating
randomness in the proposed cryptosystem. From a single system, three different chaotic
sequences are generated that are found responsible for the production of confusion and
diffusion in the three different sheets of a digital image. Performance analyses validate the
efficacy of the proposed encryption scheme. Later on in this chapter, a novel scheme of
watermarking is presented which utilizes the solution space of RF-system. Watermark is almost
impossible to sense due to the presence of chaotic sequences. Moreover, the outcomes of the
robustness test and statistical analyses support our claim that proposed watermarking technique
is semi-fragile in copyright protection.
76
Chapter 5
Lightweight Secure Image Encryption Scheme Based
on Second Order Chaotic Differential
Equation
The invention of AES in recent past is indeed a remarkable achievement in the field of
secure communication. Meanwhile, It is difficult to implement AES in constrained situations
like RFID and image encryption. In this chapter, a chaotic oscillator generated by a second
order differential equation is used to produce confusion and diffusion in the plaintext message
to achieve the desired secrecy. The produced chaotic sequence of random numbers from
dynamical system is utilized to scramble the pixels of an image to obtain an encrypted image.
Chaos based encryption technique is found secure enough to tackle chosen plaintext attacks
and brute force attacks. The specific attributes of chaotic system like, sensitivity to initial
conditions, randomness and uncertainty make it suitable for the design of cryptosystem. The
dominance of the proposed scheme is acknowledged due to the fact of better cryptographic
properties when compared with the algorithms developed already in the literature.
5.1 Introduction
Recent developments in the field of technology changed the total communication
scenario of the globe. A widespread placements of s mall computing devices is one of the major
defining trend of this century. These devices are not only used by the individual consumers but
itโs also becoming necessity of different global organizations. This advancements made the
77
general public to relay upon different wired and wireless channels for communication. The
official as well private affairs of organizations and individuals are transmitted as a matter of
correspondence.
Some of the data is of sensitive nature which always demand a secure mode of
transmission. An established fact of such computing devices is the security risk associated to
its userโs data. These threats include protection of data, illegitimate access of data and the
authenticity of the data during transmission are some of the prime concerns [67]. The excessive
deployment of such devices and security risks attached with them motivate cryptographers to
design a mechanism for the safe transmission of an individualโs data. This goal can be achieved
by developing secure algorithm for information security.
The communication by the means of images is increasing exponentially in private as
well as in official affairs. Many official correspondence sometimes utilizes images for fast
communication of the notifications and office orders. This correspondence is from the all
sectors of community including financial transactions, medical reports and military movements
etc. The images communicated without secure channel are always under the threat of
information loss and theft. The importance of secure image encryption is indeed the necessity
of this era. This application scenario motivated researchers to design image encryption
techniques [68].
Chaos and image encryption are closely linked after the theory of Shannon [1]. The
features like sensitivity towards parameters and initial conditions, randomness and ergodicity
of a chaotic system make them fit in cyber security. A cryptosystem utilizing chaos is difficult
to break as compared with a linear system or a nonlinear system without chaotic trajectories.
Different analysis for image encryption specifically motivated researchers to use chaotic
system. Chaos based system has one more advantage of reducing rounds of encryption that
78
reduces the complexity of the algorithms while maintaining the security of the system up to the
mark [20].
Ciphers are broadly divided into two categories i.e. symmetric key ciphers and
asymmetric key ciphers. The former is used in this article which has additional two principal
partitions i.e. stream ciphers and block ciphers. In block ciphers, the original information is
classified into the blocks of equal sizes. The block encryption phases (key addition, mixing,
substitution and permutation) [2] are then applied on these blocks to convert them into non
readable form. Confusion and diffusion are the two most imperative goals for cryptographers
to seek in block ciphers [20]. The objective of confusion is to assure that key should not depend
simply on cipher text i.e. each part of the enciphered data should be contingent to the several
parts of the key. Whereas the diffusion is responsible for the alteration of several output bits
whenever a single input bit is changed. These two traits in different cryptosystems are usually
attained with the help of several rounds of encryption.
Wang et.al proposed an image encryption scheme using CML and DNA sequences
operations in [40]. At first, image is encrypted using DNA code to make a DNA matrix,
afterwards which is scrambled. Lastly, scrambled matrix is diffused via CML and DNA
sequences to generate an enciphered image. Likewise in [39] Turan et.al offered an encryption
scheme for images based on total diffusion. They utilised logistic map to generate random
sequences for its further use in image encryption scheme. Similar work based on chaos and
DNA was used by Mondal et.al in [48] to design lightweight image privacy scheme. Chai et.al
in [50] suggested an image encryption scheme based on 2D logistic map and DNA sequences.
Image encryption scheme based on hyper chaotic system was proposed by Ahmed et.al
in [43]. Attaullah et.al in [41] suggested an image privacy preserving scheme based on S-box
and chaotic system. A chaotic system of dimension three was used by Naseer et.al in [42] to
design image encryption scheme. In [45], authors proposed the concept of substitution,
79
permutation and diffusion for the encryption of data after generating chaotic sequences from
chaotic maps based on beta functions. Belazi et.al coined the idea of linear fractional
transformation and lifting wavelet transform for the image encryption in [46]. Two dimensional
cat map along with the idea of bit level permutation was proposed for the encryption scheme
by Safwan et.al in [47]. This idea was further extended by Zhang et.al in [49] by using three
dimensional cat map and Chen system to suggest three dimensional bit matrix rearrangement
for image privacy scheme. Chaotic substitution box was used for image encryption in [51] by
Cavusoglu et.al. Complex chaotic system and DNA sequences was used for image encryption
algorithm by Li et.al in [38]. Moysis et.al applied the knowledge of 2D chaotic maps for the
encryption scheme [53]. Similar chaos-based encryption techniques are constructed by
researchers in [56] and [58].
Randomness produced by chaotic Lorenz system was used for the generation of S-
boxes in [69] by Khan et.al and in [70] by Ozkaynak et.al. Khan et.al in [15] suggested an idea
for the construction of S-boxes based on Henon chaotic map and symmetric group. Razzaq
et.al in [71] utilized bijective map and coset diagram for the generation robust and secure S-
boxes. Attaullah et.al in [8] presented an algorithm to design S-box based on chaotic maps after
improving chaotic range of chaotic maps.
5.2 Chaotic Duffing Oscillator
The simplest and complex systems generating chaos are logistic map and partial differential
equations respectively, while the latter one is difficult to analyse analytically and numerically.
There is a route to chaos in between these two which is built on ordinary differential equation.
The nonlinear second order differential equation that is basically used to model certain damped
and driven oscillators [19] is given as:
80
Where ๐ฅ is representing the displacement vector, whose first and second derivative
denotes the velocity and acceleration respectively. The constants ๐ผ, ๐ฝ, ๐พ and ๐ are to decide the
chaotic behavior.
๐2๐ฅ
๐๐ก2+ ๐ผ
๐๐ฅ
๐๐ก+ ๐ฝ๐ฅ + ๐พ๐ฅ3 = ๐ cos(๐๐ก) (5.1)
Fig 1. Phase space, Poincare section and time series analysis of Duffing Oscillator
Usually a forcing term and term for friction is altered and found responsible for the chaotic
nature of the Eq. (5.1). Some of the variation in both these terms are given in the Eq. (5.2) and
its bifurcation pattern, time series and Poincare section are plotted as well.
๐2๐ฅ
๐๐ก2+ ๐ผ
๐๐ฅ
๐๐ก+ ๐ฝ๐ฅ + ๐พ๐ฅ3 = ๐(๐โ๐๐๐๐๐โ๐ก + cos(๐๐ก)) (5.2)
๐ฟ๐๐ก ๐๐ฅ
๐๐ก= ๐
๐๐
๐๐ก+ ๐ผ๐ + ๐ฝ๐ฅ + ๐พ๐ฅ3 = ๐(๐โ๐๐๐๐๐โ๐ก + cos(๐๐ก)) (5.3)
-3 -2 -1 0 1 2 3
-3
-2
-1
0
1
2
3
phase space
80 100 120 140 160 180
-2
-1
0
1
2
3
time series
-3 -2 -1 0 1 2
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Poincar? section
80 100 120 140 160 180
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
time series
81
Fig 2. Phase space, Poincare section and time series analysis of Duffing Oscillator by slightly varying forcing factor
The complex and chaotic behaviour visible by the phase space diagram is being strengthen by
the periodicity of dynamical system. Such chaotic, damped and driven oscillators might be a
problem in real word physics but its randomness is found helpful in the design of nonlinear
component of block ciphers.
5.3 Substitution Box Based on Chaotic Dynamical System
There are mainly three steps involved in the algorithm for the construction of
substitution box. Initially, we have solved the chaotic system (3) numerically to obtain numeric
outcomes. Secondly, we transformed these numeric values into the integer values ranging in
the set {0,1,2, โฆ 255}. Lastly, we gathered the distinct 256 values and discarded the remaining
repeated values.
Algorithm
P.1: Suppose the obtained numerical solution of the chaotic system (3) is ๐ = {๐ข1, ๐ข2, ๐ข3, โฆ , ๐ข๐} โ ๐ .
P.2: Let ๐ be the vector such that ๐ = ๐๐๐๐[(๐ข1, ๐ข2, ๐ข3, โฆ , ๐ข๐) ร 10000] โ ๐ .
P.3: Obtain vector ๐ such that ๐ = ๐๐๐(๐, 256) โ ๐บ๐น(28).
-3 -2 -1 0 1 2 3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
phase space
80 100 120 140 160 180
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5time series
-3 -2 -1 0 1 2 3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Poincar? section
82
P.4: Extract distinct first 256 integers from the set ๐ to generate chaotic S-boxes.
Table 1: Proposed substitution box.
123 131 87 199 57 165 157 47 172 132 205 139 203 116 97 146
162 149 81 77 183 5 202 108 55 4 75 128 170 174 41 222
148 10 155 62 15 63 229 201 140 53 67 236 215 107 100 136
239 32 211 69 198 89 42 35 160 135 30 184 210 253 96 25
252 232 235 43 61 247 34 228 185 204 85 141 158 175 194 111
191 200 238 245 147 251 23 152 180 13 168 143 193 6 125 117
83 197 177 106 98 79 166 224 231 169 50 218 178 124 119 206
130 192 227 48 219 16 86 241 74 233 60 88 31 84 8 156
94 54 9 12 186 28 250 196 37 65 151 246 213 93 52 225
255 101 137 127 92 216 189 38 26 164 159 33 248 36 113 45
190 214 17 134 40 59 20 72 22 120 110 44 179 68 0 161
249 64 80 3 51 223 105 114 19 118 29 181 82 244 230 91
243 2 39 121 240 11 58 182 109 90 122 221 173 78 138 209
167 70 187 171 14 234 226 126 27 49 95 66 7 76 154 115
46 153 133 150 254 163 208 99 73 24 129 176 112 18 212 71
237 102 242 21 207 103 144 1 220 145 104 195 188 217 142 56
5.4 Algebraic Strength of an S-box
The strength of an S-box is a decisive tool sanctioning its further application in
algorithms of encryption. The importance of an S-box in any encryption standard is due to the
fact that it is the only nonlinear component of block ciphers. After the invention of AES many
techniques have been articulated in the literature for the proposal of an S-box construction.
This contest is to achieve the optimal S-box based on its properties. Different algebraic, chaotic,
linear and nonlinear systems are being addressed for the construction of S-box. This is not over
yet because the target is not achieved up till now. Loopholes like low dimensional systems,
small key space, unable to resist against linear attacks are the most prominent problems related
83
to the aforementioned S-boxes. These drawbacks need novel and complex chaotic dynamical
system to enhance the strength of an S-box.
5.4.1 Nonlinearity
Consider a set ๐ถ which is the collection of all affine transformations. The distance of any ๐ โ
๐ถ from all the elements of ๐ถ is measured in nonlinearity NL analysis. This highlights the
alteration in the truth table of Boolean functions to get nearer to the neighboring affine
transformation. In other words, the minimum hamming distance of ๐ โ ๐ถ from a set ๐ถ gives
nonlinearity of ๐. The average value of NL for the proposed S-box is higher than Ref [35], [34],
[8], [32] and [36].
5.4.2 Bit Independence Criterion
One of the imperative criterion for cryptographic structures is bit independence
criterion (BIC). The output bits independency as a consequence of exclusive input alterations
is measured in BIC [37]. The properties like confusion and diffusion can be maximized by
enhancing independence between the bits. The range of BIC lies in [0, 1] whose least value is
considered as optimal while maximum of this interval is said to be worst value of BIC.
99 1
00
10
6
10
2
10
2 10
6
10
6
10
6 10
8 10
8
10
8
11
0
10
3.3
10
3
10
6.7
10
4.7
10
5.2 1
07
.5
R E F . [ 3 5 ] R E F . [ 3 4 ] R E F . [ 8 ] R E F . [ 3 2 ] R E F . [ 3 6 ] P R O P O S E D
NONLINEARITY COMPARISON
Minimum Maximum Average
84
Table 2: Comparison of BIC-SAC values with some related work.
S-boxes Proposed Ref. [35] Ref. [34] Ref. [8] Ref. [32] Ref. [36]
BIC-SAC 0.5049 0.4987 0.5050 0.4989 0.5058 0.4956
BIC NL 104.643 104.2 106.6 106.6 104.1 103.8
5.4.3 Strict Avalanche Criterion
Consider ๐: ๐2๐ โ ๐2
๐, SAC is said to be satisfied for ๐ if for every ๐ (1 โค ๐ โค ๐), the following
equation holds
โ ๐(๐ข)โ ๐(๐ข โ ๐ถ๐๐) = (2๐โ1, 2๐โ1, โฆ , 2๐โ1)
๐ขโ๐2๐
(5.4)
This criterion is satisfied easily for a strong Boolean S-box. This implies that the chance of
alteration in output bits should be one half whenever single input bit is complemented.
Table 3: Strict avalanche criterion for the suggested S-box.
S-boxes Proposed Ref. [35] Ref. [34] Ref. [8] Ref. [32] Ref. [36]
Minimum 0.4062 0.4140 0.4218 0.4219 0.3906 0.4297
Maximum 0.5781 0.6015 0.6093 0.5625 0.5937 0.5313
Average 0.4997 0.499 0.500 0.4939 0.506 0.496
5.4.4 Differential Approximation Probability
The differential homogeneity validated by S-box is tested with the help of differential
approximation probability. This is measured by scrutinizing each single input bit for the
affirmation that identical mapping is used. Mathematically,
๐ท๐๐ (โ๐ข โ โ๐ฃ) =[#{๐ข โ ๐\๐(๐) โ ๐(๐ข โโณ ๐ข) =โณ ๐ฃ}]
2๐ (5.5)
The fallouts of probabilities of disparity by applying input and output discrepancies are given
in Table 5.
Table [5]. Evaluation table for DP outcomes
S-boxes Proposed Ref. [35] Ref. [34] Ref. [8] Ref. [32] Ref. [36]
85
Max. DP 0.04687 0.03906 0.05469 0.0625 0.04688 0.03906
5.4.5 Linear Approximation Probability
The Let ๐ถ is a set of all probable inputs whose total elements are 2๐. If the parity of
input and output bits is denoted by ฮ๐ โ 0 ๐๐๐ ฮ๐ โ 0 respectively, then linear approximation
probability is given by the following equation
๐ฟ๐ = ๐๐๐ฅ |{โ๐ง๐งโฮ๐ = ๐(๐ง)โฮ๐}
2๐โ1
2| (5.6)
It calculates the maximum worth of imbalance of an outcome or the disproportion of
an event. Linear approximation probability of the suggested S-box is 0.1484. Comparison of
LP with different S-boxes is given in Table 4.
Table 6: Comparison table of LP results for numerous S-boxes
S-box Suggested Ref. [35] Ref. [34] Ref. [8] Ref. [32] Ref. [36]
Max. LP 0.1484 0.1328 0.1289 0.1250 0.1250 0.1562
Max.
Value
164 164 162 160 160 168
5.5 Proposed Scheme for Image Encryption
This section introduces an encryption algorithm which utilizes both i.e. the numerical
solution and substitution box generated by new attractor defined in section 3. The whole
procedure is described in the following steps.
Rotation: Initially, an original image ๐(๐, ๐) of dimension ๐ ร๐ is rotated clockwise by 90ยฐ.
This rotation is explained as follow
๐โฒ(๐, ๐) = ๐(๐ โ ๐ + 1, ๐) (5.7)
Substitution: In this phase, pixels of a rotated image are substituted by the entries of
substitution box. This muddles the image ๐(๐ข, ๐ฃ) to an extreme extent generating a total
86
meaningless image ๐โฒ(๐ข, ๐ฃ). Rijndael substitution process is used to substitute the pixel value
with S-box entries i.e. most significant bits MSBs and least significant bits LSBs are obtained
after converting the pixel values into binary. Former represent the column of an S-box and
latter denotes the row of an S-box. The intersection of these two gives an entry which is to be
replaced with that pixel value.
Row Extraction: Row matrices ๐ ๐ข are extracted from the above matrix ๐โฒ(๐ข, ๐ฃ) having ๐ฃ =
๐ + 1 columns.
Chaotic Swapping: Each row matrix obtained above is substituted by the solution of Duffing
equation and then all rows are concatenated to obtain a matrix. The added column is then
discarded to yield the encrypted image ๐ธ(๐ , ๐ก) of first round. The same procedure is repeated
four times to generate resulting encrypted image.
The data values excluding the first column for each ๐ ๐ข are switched with the numerical solution
of Duffing Equation, the whole procedure is given as:
๐(๐) = ๐๐(๐ โ 1) โจ ๐ ๐(๐)โจ (โ๐๐(๐, ๐) ร 1010โ๐๐๐256) ๐๐๐ 1 < ๐ โค (๐พ + 1) (5.8)
where โโโ denotes ceiling function. For each cycle (๐ = 1,2,3,4) encryption procedure ๐๐(๐, ๐)
is represented by the following equation:
๐๐(๐, ๐) = {
๐๐โ1(๐พ, 0) ๐ = 0, ๐ = 2,4, ๐ = 0
โ๐ท๐ข๐๐๐๐๐(๐ผ๐, ๐๐(๐ โ 1,0)) ๐ = 0, ๐ > 0
โ๐ท๐ข๐๐๐๐๐(๐ผ๐, ๐๐(๐, ๐ โ 1)) ๐ > 0, ๐ > 0
(5.9)
here ๐ผ๐ and ๐๐(0,0) denotes the constraint for Duffing equation and the primary condition for
the ๐๐กโ enciphering round respectively. Customer chooses the set of security keys involving
๐1(0,0), ๐ผ0 and ๐ผ๐.
87
Original Image
Clockwise Rotation
Addition of Random
Column
S-box Substitution utilizing Rijndal Transformation
Separate the 2D image Into
(M+1) Row Matrices
Replace each entry of the row
matrix with the Numerical
Solution of Chaotic Oscillator
Reunite all the Row
Matrices
Encrypted Image
Discard the First
Column of the 2D
Matrix
Fig. 3. Suggested encryption scheme.
Fig 4. Original and Encrypted Lena Image of dimension 256 ร 256
(a) (b) (c)
Fig 5. Layer-wise original images of Lena (a)-Red, (b)-Green, (c)-Blue
(d) (e) (f)
Fig 6. Layer-wise Encrypted images of Lena (d)-Red Encrypted, (e)-Green Encrypted, (f)-Blue Encrypted
88
Fig 7. Original and Encrypted Baboon Image of size 512 ร 512.
5.6 Investigational Upshots and Simulation Analyses
In any investigation of designed cryptosystems the ultimate gauge is to measure the
outcomes of different analyses. The astonishing fact connected to any research is the disclosure
of false outcomes after a long and hectic tiresome job. Sometimes, for scientist and engineers,
it become really hard to identify the wrong step. Still itโs an interesting task for many. The
efficacy of any scheme is established right after the complete investigation of analyses. For
this, simulation analyses of proposed scheme are given hereafter.
5.6.1 Key Space Analysis
In the design of a cryptosystem the overall number of various keys used generates the
key space. The strength of a cryptosystem against brute force attack is directly linked with the
key space i.e. larger the key space implies the more secure the proposed cryptosystem.
Generally, the key space for chaos based cryptosystem must be greater than 2100 [72]. There
are six total different keys that are being used in the solution of differential equation to obtain
chaotic trajectory. Hence in view of [57], the total key space is 1015ร6 โ 2280, which is much
superior that key space of [73] whose key space is 2128. Thus the key space is enough for the
confrontation of brute force attacks.
89
5.6.2 Key Sensitivity Analysis
A prominent characteristic of chaos based system is its sensitivity towards parameters
and initial input. Whenever initial conditions and parameters are slightly altered, the output
behaviour of chaotic system changes abruptly. Observation of such dissimilar dynamics for a
single mathematical problem makes reader difficult to accept it as a one system. Such sensitive
attribute of a dynamical system is best suited for image encryption. Any attempt by an
illegitimate user to decrypt the message will generate the wrong original message. This surety
is just because of sensitivity of the system towards keys. The proposed encryption technique
as well as S-box are designed with the help of chaotic system which is highly sensitive towards
the initial inputs that successfully clear the key sensitivity test.
5.6.3 Correlation Analysis
The pixels of an actual image horizontally, vertically and diagonally are highly
correlated. The maximum value in either direction is around one. Whereas the correlation of
an enciphered image must be approaching to zero which confirms the efficacy of the
cryptosystem. Any designed cryptosystem tends to distort/break the relationship between the
neighbouring pixels in such a way that no one can refigure the original status of the pixels. The
correlation values in all the three directions are tabulated in Table 6.
Table [6]. Correlation outcomes in all directions
Image Layer Horizontal Diagonal Vertical
Original Baboon
Image
R 0.9414 0.9006 0.9228
G 0.9033 0.8203 0.8842
B 0.9438 0.9059 0.9331
Encrypted Baboon
Image by Proposed
Scheme
R 1.0563 ร 10โ5 โ6.2038 ร 10โ5 7.2365 ร 10โ5
G โ4.0134 ร 10โ5 5.1344 ร 10โ5 โ4.1006 ร 10โ5
B โ1.7421 ร 10โ5 1.8515 ร 10โ5 2.3056 ร 10โ5
R โ0.0072 โ0.0016 โ0.0201
90
Encrypted Baboon
Image by Ref [36]
G โ0.0260 โ0.0175 โ0.0220
B โ0.0099 โ0.0066 โ0.0034
Ref [59] 0.0038 4.001 ร 10โ4 0.0023
Ref [74] โ8.3729 ร 10โ4 โ9.0317 ร 10โ4 0.0011
5.6.4 Histogram Analysis
The graphical illustration of the tonal distribution in an image gives the histogram of
an image. This can be achieved by plotting number of pixels in a particular tone along y-axis
for every single tonal value along x-axis. Lighter areas are represented towards right side of
the horizontal axis while darker areas come along the other side. In other words, histogram of
dark image will lie towards left and middle of the x-axis.
The histogram of an original image gives the actual distribution of pixel values and its
graph is normally randomly distributed as per the real image pixels dispersal. Whereas, the
histogram of an enciphered image is flat if it is encrypted with a sound encryption scheme. The
histogram of the proposed scheme are shown below.
Fig 8. Histogram of original Lena image
(f). Red Image (original) (g). Green Image (original) (h). Blue Image (original)
Fig 9. Layer wise view of Lena image (f)-(h).
0 50 100 150 200 250 3000
200
400
600
800
1000
1200
1400
91
(i)- Histogram of (f) (j)- Histogram of (g) (k)- Histogram of (h)
Fig 9. Layer wise view of histogram of Lena image (i)-(k).
(l)-Encrypted Red Image (m)-Encrypted Green Image (n)-Encrypted Blue Image
Fig 10. Layer wise encrypted Lena image (l)-(m).
(o)-Histogram of (l) (p)-Histogram of (m) (q)-Histogram of (n)
Fig 11. Layer wise histogram of encrypted Lena image (o)-(q).
Fig. 12. Plaintext image and Histogram of the original Baboon image
0
100
200
300
400
500
600
700
800
900
1000
0 50 100 150 200 250
0
100
200
300
400
500
600
700
0 50 100 150 200 250
0
100
200
300
400
500
600
700
800
0 50 100 150 200 250
0
100
200
300
400
500
600
0 50 100 150 200 250
0
100
200
300
400
500
600
0 50 100 150 200 250
0
100
200
300
400
500
600
0 50 100 150 200 250
92
Fig. 13. Encrypted Baboon image and its Histogram
5.6.5 Information Entropy
The amount of randomness appeared in an image is an important tool to decide the
efficacy of an encryption algorithm. Information entropy (IE) is used to calculate this
randomness and unpredictability. The peak value of IE for an enciphered image is 8. It implies
that histogram of an encrypted image will be uniformly distributed and hence it will be flat.
The information entropy analysis for the proposed encryption algorithm is given in Table 7.
The values of IE are close to 8 implying the robustness against entropy attacks. The
mathematical formulation for the measurement of IE ๐ป(๐ข๐) of a random variable (r.v) ๐ข๐ is
given by the following relation.
๐ผ๐ธ(๐ป) = โโ๐(๐ข๐)๐๐๐2๐(๐ข๐) (5.10)
Where ๐(๐ข๐) denotes the probability of a r.v ๐ข at ๐๐กโ index.
Table [7]: Comparison of information entropy.
Imageries Layer Suggested Ref. [36] Ref. [58] Ref. [7]
Baboon Image R 7.9978 7.9987 7.9981 7.9973
G 7.9988 7.9989 7.9974 7.9969
B 7.9987 7.9990 7.9977 7.9985
Pepper Image R 7.9989 7.9992 7.9986 7.9972
G 7.9990 7.9991 7.9972 7.9977
B 7.9972 7.9988 7.9968 7.9987
Lena Image R 7.9891 7.9882 7.9981 7.9893
G 7.9894 7.9890 7.9982 7.9882
B 7.9895 7.9888 7.9889 7.9885
5.6.6 Sensitivity Analysis
To recover encrypted image, cryptanalyst make use of chosen plaintext attack, also
known as differential attack. A good cryptosystem resisting differential attack must be design
93
in such a way that a small modification in the host image will generate a totally different
encrypted image. For the measurement of this attribute of a cryptosystem, (NPCR) number of
pixels change rate and (UACI) unified average changing intensity are utilized usually for the
differential analysis [75]
๐๐ด๐ถ๐ผ(๐ถ1, ๐ถ2) =1
๐ ร ๐ป| ๐ถ1(๐, ๐) โ ๐ถ2(๐, ๐)
255| ร 100% (5.11)
๐๐๐ถ๐ =โ ๐ท(๐, ๐)๐,๐
๐ ร ๐ปร 100% (5.12)
Where,
๐ท(๐, ๐) = {0 ๐๐ ๐ถ1(๐, ๐) = ๐ถ2(๐, ๐)
1 ๐๐ ๐ถ1(๐, ๐) โ ๐ถ2(๐, ๐)} (5.13)
Table 8. Valuation of UACI outcomes for proposed scheme.
Image Layer Proposed Ref. [36] Ref. [58] Ref. [7]
Baboon Coloured Image R 33.51 % 33.54 % 33.65 % 33.75 %
G 33.59 % 33.61 % 33.59 % 33.68 %
B 33.66 % 33.64 % 33.67 % 33.60 %
Pepper coloured Image R 33.60 % 33.54 % 33.66 % 33.68 %
G 33.59 % 33.65 % 33.66 % 33.54 %
B 33.63 % 33.62 % 33.73 % 33.65 %
Lena Coloured Image R 33.63% 33.51% 33.62% 33.61%
G 33.51% 33.55% 33.64% 33.66%
B 33.57% 33.67% 33.79% 33.74%
Table [9]. Comparison table of NPCR fallouts of suggested scheme.
Image Layer Ref. [36] Ref. [58] Ref. [7] Proposed
Baboon Image R 99.61 % 99.62 % 99.63 % 99.57 %
G 99.62 % 99.63 % 99.63 % 99.61 %
B 99.61 % 99.63 % 99.62 % 99.62 %
Lena Coloured Image R 99.64 % 99.59 % 99.60 % 99.63%
G 99.61 % 99.60 % 99.62 % 99.64%
B 99.58 % 99.64 % 99.61 % 99.60%
Pepper coloured Image R 99.62 % 99.62 % 99.62 % 99.66 %
G 99.62 % 99.62 % 99.62 % 99.61 %
94
B 99.62 % 99.62 % 99.63 % 99.60 %
5.7 Randomness Test (NIST SP 800-22) for cipher
NIST SP 800-22 [26] is a statistical test used to measure random and pseudorandom number
generators for cryptographic applications. For this purpose, a baboon image of 512 ร 512 is
encrypted utilizing the proposed scheme given in section 5, the enciphered image is passed
through NIST SP 800-22 to clear the test for authentication of randomness. The outcomes like
long runs of ones, block frequency, and random excursion test etc. are passed by the encrypted
image. Tests results of NIST are tabulated in table [10].
Fig 14: Original and Encryption Baboon Image of dimension 512 ร 512
Table 10: NIST test results for Encrypted RGB Baboon Image
Test P โ values for colour encryptions of ciphered image
Results
Red Green Blue
Frequency 0.15443 0.12136 0.63626 Pass
Block frequency 0.7267 0.25453 0.57678 Pass
Long runs of ones 0.7137 0.7137 0.7137 Pass
Overlapping templates 0.85888 0.81767 0.85888 Pass
No overlapping templates 0.99971 1 0.96015 Pass
Approximate entropy 0.022769 0.65995 0.17905 Pass
Serial p values 1 7.2376e-06 0.30904 0.11831 Pass
95
In this chapter, a second order differential equation (ODE) generating chaotic solution
is utilized for the encryption of images. The parameters and factors of this ODE are slightly
varied to obtain a chaotic oscillator which is dissimilar from the well-known duffing oscillator
as for as its chaotic trajectory and range is concerned. Furthermore, the density distribution
function of this system is denser. Initially, an algorithm for the construction of substitution box
is formulated using this ODE which was found much better in analyses like nonlinearity, SAC
BIC, and differential and linear approximation probabilities. Later on, a lightweight colour
image encryption technique is proposed. The inculcation of substitution process in encryption
scheme generated diffusion in the plaintext images whereas, for the sake of confusion the
Serial p values 2 2.4617e-06 0.86074 0.36724 Pass
Cumulative sums forward 0.18502 0.28014 0.26743 Pass
Cumulative sums reverse 0.78837 1.7271 0.92583 Pass
Random excursions X = -4 0.64048 0.86804 0.98249 Pass
X = -3 0.65389 0.52779 0.97815 Pass
X = -2 0.68278 0.1896 0.96465 Pass
X = -1 0.52747 0.15759 0.8281 Pass
X = 1 0.74755 0.54947 0.093195 Pass
X = 2 0.45219 0.50705 0.98306 Pass
X = 3 0.86117 0.7266 0.99815 Pass
X = 4 0.67706 0.087546 0.98919 Pass
Random excursions variants X = -5 0.2271 0.19547 0.68209 Pass
X = -4 0.2057 0.25584 0.64943 Pass
X = -3 0.16828 0.4574 0.57988 Pass
X = -2 0.35354 0.63501 0.4785 Pass
X = -1 0.6775 0.45525 0.68309 Pass
X = 1 0.10381 0.45425 0.6809 Pass
X = 2 0.53509 0.506 0.09796 Pass
X = 3 0.58968 0.27718 0.017522 Pass
X = 4 0.38854 0.41987 0.030654 Pass
X = 5 0.50304 0.73688 0.1342 Pass
96
permutation process is being utilized. Addition of random column increases the resistance
against plaintext attacks. Moreover, the encrypted images are unpredictable, non-repeated and
random even after using the same set of initial parameters and conditions. The outcomes of
different statistical and differential analysis suggests the effectiveness of the proposed scheme.
97
Chapter 6
Conclusion and Future Directions
The whole thesis is concluded in this chapter. Moreover, some questions that arose
during this research are presented as a future work.
The main objective of this research are categorically given hereafter.
1. One dimensional and simple chaotic systems have smaller solution space and their
bifurcation pattern can be retraced with the help of strong computing systems.
2. To overwhelm this short come, chaotic dynamical systems being multi-dimensional are to
be used in the design of nonlinear component of block cipher because of their rich and
complex solution space.
3. Coupled differential equations of first order are used to obtain cryptographically strong and
large number of S-boxes.
4. Second order differential equation based chaotic oscillator is also applied in the design of
cryptosystem.
5. Finally, the aim is to utilize these systems in the invention of image encryption and
watermarking techniques.
6.1 Conclusion
The unique features of chaotic systems like sensitivity towards initial
parameters/conditions, their randomness, and complex and chaotic bifurcation pattern have
amplified their importance in information security. At the outset, a brief introduction to chaos
and cryptography is illustrated in chapter 1. Instead of using one dimensional and simple
98
chaotic maps in cyber security, the need and importance of utilization of multi-dimensional
chaotic dynamical systems is established in the same chapter. Furthermore, nonlinear
component of block ciphers and its physiognomies are discussed in detail.
In second chapter, a set of coupled differential equation named as Rabinovich-Fabrikant
(RF) system of differential equation belonging to the first order coupled differential equations
class is solved. The parameters controlling chaos in this system are altered to observe the
bifurcation pattern. Interestingly, small variation in this dynamical system yielded a totally
different chaotic trajectory. This characteristic is used to generate an S-box of qualified
cryptographic properties. Moreover, from this system one can generate many different S-boxes
using different combinations of solution sequences. Different analyses like nonlinearity, bit
independence criterion, strict avalanche criterion, and linear and differential approximation
probabilities are used to measure the strength of an S-box generated from the solution space of
RF system.
Third chapter utilizes one of simplest chaotic dynamical system i.e. double pendulum in
the design of S-boxes. The mathematical modelling of this system results in the form of set of
four differential equations. Their complexity is increased by using two double pendulums at a
time. This makes its trajectory more complex and unpredictable. The sequence of
pseudorandom numbers is collected from its solution space. Here, instead of making S-box
after this step, we have used permutation from a symmetric group to mix up the pseudorandom
numbers. This in result increased the nonlinearity of the S-box. Moreover, it also strengthen
the remaining analyses of the nonlinear component of block cipher confirming its application
in information security.
An application of RF system in the form of image encryption and watermarking is given
in chapter four. Chaotic solution of RF system with different initial conditions and parameters
is achieved using Runge-Kutta method of order four in MATLAB. Three dissimilar sequences
99
from the solution space are extracted for their further utilization in colour image encryption.
Each channel of red, green and blue are separately operated to scramble the original positioning
of pixels to yield an encrypted meaningless image. Moreover, the same sequences are
implemented for watermarking which was done along each channel. The simulation results of
all major standard analyses were observed carefully and found clear in determining the
robustness of the proposed schemes of encryption and watermarking.
In chapter five, a second order differential equation also known as Duffing equation is
debated. The chaotic behaviour of this equation is seen in the form of oscillator. The interesting
attribute of this oscillator is the production of different bifurcation pattern for slight variation
in parameters/conditions causing chaos. Its density distribution function is complex, rich and
dense. It means one can generate a large number of S-boxes from this single system.
Lightweight but yet practical and applicable image privacy preserving scheme established on
chaotic differential equation is presented in this chapter. All the standard analyses were found
promising in analysing the suggested scheme of encryption.
6.2 Future Work
Field of information security is as big as this universe. There are numerous levels that
can be addressed lonely at a one time. Many queries came across while doing the proposed
thesis which not only motivated in this but also increased the knowledge. There are still
ambiguous as well as chaotic mathematical structures that have the ability to influence
multimedia security. Some of the future research perspective work is listed hereafter.
Different Mathematical based systems should be utilized in cyber security. Like partial
differential equations, topological structures, loop theory, cost diagrams and non-
associative structures.
100
For a larger input bits, S-box must be designed for larger input bits to withstand in such
situations. Furthermore, S-box evaluation software for many input bits is also required
to be invented.
Quantum cryptography is the recent advancement in information security, in view of
this development bits are going to be replaced by cubits in near future. Their security
will be a challenge as well in coming days. Moreover, mathematical visualization of
quantum physics can also be used in the invention of new encryption standard.
With the extensive invention of soft computing devices, the data security of an
individual demands free of cost applications available on play-store and i-store for the
safe communication of general public. Scientist should work to target such populaces
by inventing apps and soft wares for the usage of an individual.
101
References
[1] C. E. Shannon, โCommunication theory of secrecy systems,โ Bell Syst. Tech. J., vol. 28,
no. 4, pp. 656โ715, 1949.
[2] J. Daemen and V. Rijmen, โThe design of Rijndael-AES: the advanced encryption
standard.,โ Springer, Berlin, 2002.
[3] J. Rehman, A. U., Khan, J. S., Ahmad, โA New Image Encryption Scheme Based on
Dynamic S-Boxes and Chaotic Maps,โ 3D Res. Springer 7, vol. 7, 2016.
[4] S. S. Jamal, T. Shah, S. Farwa, and M. U. Khan, โA robust chaotic stegano- graphic
technique with enhanced security based on a high-nonlinearity S-box. Submitted.,โ
2017.
[5] S. S. Jamal, M. U. Khan, and T. Shah, โA Watermarking Technique with Chaotic
Fractional S-Box Transformation,โ Wirel. Pers. Commun., vol. 90, no. 4, 2016.
[6] Y. Wu, Y. Gelan, H. Jin, and J. P. Noonan, โImage encryption using the two-dimensional
logistic chaotic map,โ J. Electron. Imaging, vol. 21, no. 1, p. 013014, 2012.
[7] Z. Hua, Y. Zhou, C. M. Pun, and C. L. P. Chen, โ2D Sine Logistic modulation map for
image encryption,โ Inf. Sci. (Ny)., vol. 297, pp. 80โ94, 2015.
[8] A. Ullah, S. S. Jamal, and T. Shah, โA novel construction of substitution box using a
combination of chaotic maps with improved chaotic range,โ Nonlinear Dyn., 2017.
[9] G. P. Williams, Chaos Theory Tamed. A Joseph Henry Press Book, 1997.
[10] R. May, โSimple mathematical models with very complicated dynamics,โ Nature, vol.
261, no. 5560, pp. 459โ467, 1976.
[11] N. K. Pareek, V. Patidar, and K. K. Sud, โImage encryptoin using chaotic logistic map,โ
Image Vis. Comput., vol. 24, no. 9, pp. 926โ934, 2006.
[12] R. L. Devaney, Fractal pattern arising in chaotic dynamical systems, The Science of
102
Fractal Images. 1988.
[13] M. Khan and Z. Asghar, โA novel construction of substitution box for image encryption
applications with Gingerbreadman chaotic map and S 8 permutation,โ Neural Comput.
Appl., 2016.
[14] M. Henon, โA two-dimentional mapping with a strange attractor,โ Commun. Math.
Phys., vol. 50, no. 1, pp. 69โ77, 1976.
[15] M. Khan and T. Shah, โA novel image encryption technique based on Henon chaotic
map and S 8 symmetric group,โ Neural Comput. Appl., vol. 25, pp. 1717โ1722, 2014.
[16] O. E. Rossler, โAn equation for continuous chaos,โ Phys. Lett. A, vol. 57, no. 5, pp. 397โ
398, 1976.
[17] E. N. Lorenz, โDeterministic non periodic flow,โ J. Atmos. Sci., vol. 20, no. 2, pp. 130โ
141, 1963.
[18] M. I. Rabinovich and A. L. Fabrikant, โStochastic self-mudulation of waves in non-
equilibrium media,โ J.E.T.P (sov), vol. 77, pp. 617โ629, 1979.
[19] M. J. Brennan, I. Kovacic, and A. Carrella, โOn the jump-up and jump-down frequencies
of Duffing oscillator,โ J. Sound Vib., vol. 318, pp. 1250โ1261, 2008.
[20] L. Kocarev, โchaos based Cryptography: A Brief Overview,โ Circuits Syst. Mag. IEEE,
vol. 1, no. 3, pp. 6โ21, 2001.
[21] R. Brown and L. O. Chua, โClarifying chaos: Examples and counterexamples,โ Int. J.
Bifurc. Chaos, vol. 6, no. 2, pp. 219โ249, 1996.
[22] F. Dachselt and W. Schwarz, โChaos and cryptography,โ IEEE Trans Circuits Syst, vol.
48, no. 12, pp. 1498โ509, 2001.
[23] A. F. Webster and S. Tavares, Advances in Cryptology, 85th ed. Proceedings of
CRYPTO. Lecture Notes in Computer Science, 1986.
[24] Attaullah, A. Javeed, and T. Shah, โCryptosystem techniques based on the improved
103
Chebyshev map: an application in image encryption,โ Multimed. Tools Appl., vol. 78,
no. 22, pp. 31467โ31484, 2019.
[25] I. Hussain, T. Shah, M. A. Gondal, and H. Mahmood, โGeneralized majority logic
criterion to analyze the statistical strength of s-boxes,โ Zeitschrift fur Naturforsch. - Sect.
A J. Phys. Sci., vol. 67, no. 5, pp. 282โ288, 2012.
[26] A. Rukhin et al., โA statistical test suit for random and pseudo random number
generators for cryptographic applications,โ NIST Spec. Publ. 800-22, 2001.
[27] M. Khan, T. Shah, and S. I. Batool, โConstruction of S-box based on chaotic Boolean
functions and its application in image encryption,โ Neural Comput. Appl., vol. 27, no.
3, pp. 677โ685, 2016.
[28] G. Jakimoski and L. Kocarev, โChaos and Cryptographyโฏ: Block Encryption Ciphers
Based on Chaotic Maps,โ IEEE Trans. Circuits Syst.-1 Fundam. theory Appl., vol. 48,
no. 2, pp. 163โ169, 2001.
[29] V. Daemen,J., Rijmen, โThe Design of Rijndael-AES:,โ Adv. Encry- tion Stand.
Springer Berlin ., 2002.
[30] C. P. . Zhou, Y., Bao, L., Chen, โA new 1D chaotic system for image encryption. Signal
Processing,โ vol. 97, pp. 172โ184, 2014.
[31] A. Ullah, A. Javeed, and T. Shah, โA scheme based on algebraic and chaotic structures
for the construction of substitution box,โ Multimed. Tools Appl., vol. 78, no. 22, pp.
32467โ32484, 2019.
[32] M. Khan, T. Shah, H. Mahmood, M. A. Gondal, and I. Hussain, โA novel technique for
the construction of strong S-boxes based on chaotic Lorenz systems,โ Nonlinear Dyn.,
vol. 70, no. 3, pp. 2303โ2311, 2012.
[33] M. F. Danca, N. Kuznetsov, and G. Chen, โunusual dynamics and hidden attrackters of
the Rabinovich-Fabrikant system,โ Nonlinear Dyn., vol. 88, pp. 791โ805, 2017.
104
[34] G. Chen, Y. Chen, and X. Liao, โAn extended method for obtaining S-boxes based on
three-dimensional chaotic Baker maps,โ Chaos, Solitons and Fractals, vol. 31, no. 3,
pp. 571โ579, 2007.
[35] G. Tang, X. Liao, and Y. Chen, โA Novel Method for Designing S-boxes based on
Chaotic Maps,โ Chaos, Solitons & Fractals, vol. 23, no. 2, pp. 413โ419, 2005.
[36] A. Belazi, M. Khan, A. A. A. El-Latif, and S. Belghith, โEfficient cryptosystem
approaches.pdf,โ Nonlinear Dyn., vol. 87, no. 1, pp. 337โ361, 2017.
[37] A. F. Webster and S. Tavares, โOn the design of S-boxes. In: Advances in Cryptology,
Lecture Notes in Computer Science,โ in Proceedings of CRYPTOโ85, 1986, pp. 523โ
534.
[38] X. Li, L. Wang, Y. Yan, and P. Liu, โAn improvement color image encryption algorithm
based on DNA operations and real and complex chaotic systems,โ Optik (Stuttg)., vol.
127, no. 5, pp. 2558โ2565, 2016.
[39] O. Jakub, J. Turan, and L. Ovsenik, โAn image encryption algorithm with total
diffusion,โ Carpathian J. Electron. Comput. Eng., vol. 11, no. 1, pp. 15โ25, 2018.
[40] X. Wang, Y. Hou, S. Wang, and R. Li, โA new image encryption algorithm based on
CML and DNA sequence,โ IEEE access, vol. 10.1109/ac, 2018.
[41] A. Ullah, S. S. Jamal, and T. Shah, โA novel scheme for image encryption using
substitution box and chaotic system,โ Nonlinear Dyn., vol. 91, no. 1, pp. 359โ370, 2018.
[42] Y. Naseer, D. Shah, and T. Shah, โA novel approach to improve multimedia security
utilizing 3D mixed chaotic map,โ Microprocess. microsystems, vol. DOI.org/10, 2018.
[43] M. Ahmed, M. N. Doja, and S. M. N. Beg, โSecurity analysis and enhancements of an
image cryptosystem based on hyper chaotic system,โ J. King saud Univ. Comput. Inf.
Sci., vol. DOI.org/10, 2018.
[44] H. M. Waseem, M. Khan, and T. Shah, โImage privacy scheme using quantum spinning
105
and rotation,โ Electron Imaging, vol. 27, no. 06, 2018.
[45] R. Zahmoul, R. Ejabli, and M. Zaied, โImage encryption based on new Beta chaotic
map,โ Opt. Lasers Eng., vol. 96, pp. 39โ49, 2017.
[46] A. Belazi, A. Ahmed, A. Diaconu, R. Rhouma, and S. Belghith, โChaos based partial
image encryption scheme based on linear fractional and lifting wavelet transform,โ Opt.
Lasers Eng., vol. 88, pp. 37โ50, 2017.
[47] S. E. Assad and M. Farajallah, โA new chaos-based image encryption system,โ Signal
Process. Image Commun., vol. 41, 2016.
[48] B. Mondal and T. Mandal, โA light weight secure image encryption scheme based on
chaos and DNA computing,โ J. King saud Univ. Comput. Inf. Sci., vol. 29, pp. 499โ504,
2017.
[49] W. Zhang, H. Yu, Y. L. Zhao, and Z. L. Zhu, โImage encryption based on three-
dimensional bit matrix permutation,โ Signal Processing, vol. 118, pp. 36โ50, 2016.
[50] X. Chai, Y. Chen, and L. Broyde, โA novel chaos-based image encryption algorithm
using DNA sequence operations,โ Opt. Lasers Eng., vol. 88, pp. 197โ213, 2017.
[51] U. Cavusoglua, S. Kacar, I. Pehlivan, and A. Zengin, โSecure image encryption
algorithm design using a novel chaos-based,โ Chaos, Solitons and Fractals, vol. 95, pp.
92โ101, 2017.
[52] X. Li, L. Wang, Y. Yan, and P. Liu, โAn improvement color image encrytion algorithm
based on DNA operations and real and complex chaotic system,โ Opt-Int J. Light
Electron opt, vol. 127, no. 5, pp. 2558โ2565, 2016.
[53] L. Moysis and A. T. Azar, โNew discrete time 2D chaotic maps,โ Int J Syst Dyn Appl,
vol. 6, no. 1, pp. 77โ104, 2017.
[54] H. Liua and X. Wang, โColor image encryption based on one-time keys and robust
chaotic maps,โ Comput. Math. with Appl., vol. 59, no. 10, pp. 3320โ3327, 2010.
106
[55] A. Anees, โAn Image Encryption Scheme based on Lorenz System for low profile
applications,โ 3D Res., vol. 6, no. 3, pp. 6โ24, 2015.
[56] X. Wang, L. Teng, and X. Qin, โA novel colour image encryption algorithm based on
chaos,โ Signal Processing, vol. 92, no. 4, pp. 1101โ1108, 2012.
[57] โIEEE Standard for binary floating point arithmatic,โ IEEE Comput. Soc., 1985.
[58] X. Wang, L. Liu, and Y. Zhang, โA novel chaotic block image encryption algorithm
based on dynamic random growth technique,โ Opt. Lasers Eng., vol. 66, pp. 10โ18,
2015.
[59] S. Behnia, A. Akhshani, H. Mahmodi, and A. Akhavan, โA novel algorithm for image
encryption based on mixture of chaotic maps,โ Chaos, Solitons and Fractals, vol. 35,
no. 2, pp. 408โ419, 2008.
[60] I. Hussain, T. Shah, and M. Asif, โAn efficient image encryption algorithm based on S
8 S-box transformation and NCA map,โ Opt. Commun., vol. 285, no. 24, pp. 4887โ4890,
2012.
[61] D. P. Mukherjee, S. Maitra, and S. T. Acton, โSpatial Domain Digital Watermarking of
multimedia objects for Buyer Authentication.,โ IEEE Trans. Multimed., vol. 6, no. 1, p.
2004.
[62] S. D. Lin and C. F. Chen, โA robust DCT-based watermarking for copyright protection,โ
IEEE Trans. Consum. Electron., vol. 46, pp. 415โ421, 2000.
[63] G. Caronni, โAssuring ownership rights for digital images,โ in Proceedings of Reliable
IT Systems, viewveg Publishing Company, Germany, 1995, pp. 251โ263.
[64] Q. Su, G. Wang, X. Zhang, G. Lv, and B. Chen, โA new algorithm of blind color image
watermarking based on LU decomposition,โ Multidimens. syst signal Process, 2017.
[65] Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, โImage qualty assessment:
From error visibilty to structural similarity,โ IEEE trans image Process, vol. 13, no. 4,
107
pp. 600โ612, 2004.
[66] U. Cox, L. Kilian, F. Leighton, and T. Shamoon, โSecure spread spectrum watermarking
for multimedia,โ IEEE Trans. Image Process., vol. 6, pp. l673โ1687, 1997.
[67] C. Paar and J. Pelzl, Understanding cryptographuy: A text book for students and
practitioners. New York: Springer Verlag, 2010.
[68] K. T. Alligood, T. D. Sauer, and J. A. Yorke, Chaos: An introduction to dynamical
system. New York: Springer Verlag, 1996.
[69] M. Khan, T. Shah, and H. Mahmood, โAn efficient method for the construction of block
cipher with multi-chaotic systems,โ Nonlinear Dyn., vol. 71, pp. 489โ492, 2013.
[70] F. Ozkaynak and A. B. Ozer, โA method for designing strong S-boxes based on chaotic
Lorenz system,โ Phys. Lett. A374(36), 3733-3738 (2010)., vol. 374, no. 36, pp. 3733โ
3738, 2010.
[71] A. Razaq, A. Yousaf, U. Shuaib, N. Siddiqui, A. Ullah, and A. Waheed, โA Novel
Construction of Substitution Box Involving Coset Diagram and a Bijective Map,โ Secur.
Commun. networks, 2017.
[72] G. Alvarez and S. Li, โSome basic cryptographic requirements for chaos based
cryptosystem,โ Int. J. Bifurcat. Chaos, pp. 2129โ2151, 2006.
[73] X. Y. Wang, L. Yang, R. Liu, and A. Kadir, โA chaotic image encryption algorithm
based on perception model,โ Nonlinear Dyn., vol. 62, no. 3, pp. 615โ621, 2010.
[74] X. Liao, S. Lai, and Q. Zhou, โA novel image encryption algorithm based on self-
adaptive wave transmission,โ Signal Processing, vol. 90, no. 9, pp. 2714โ2722, 2010.
[75] Y. Wu, J. P. Noonan, and S. Agaian, โNpcr and uaci randomness tests for image
encryption,โ Cyber J Multidisc J Sci Technol J Sel Are Telecommun, 2011.