Esteganografia tesis-master

High Capacity Irreversible Image Steganography A thesis submitted in partial fulfilment of the requirements for the award of the degree of

Master of Technology in

Computer Science and Engineering By Soumendu Chakraborty Roll No:108151010

Department of Computer Engineering & Applications Institute of Engineering & Technology

GLA University Mathura- 281406, INDIA July, 2013

Department of computer Engineering and Applications GLA University, Mathura 17 km. Stone Nh#2, Mathura-Delhi Road, P.O. – Chaumuha, Mathura – 281406 U.P (India)

Declaration I hereby declare that the work which is being presented in the M.Tech. thesis “High Capacity Irreversible Image Steganography”, in partial fulfillment of the requirements for the award of the Master of Technology in Computer Science and Engineering and submitted to the Department of Computer Engineering and Applications of GLA University, Mathura, is an authentic record of my own work carried under the supervision of Dr. Anand Singh Jalal, Associate Professor. The contents of this thesis, in full or in parts, have not been submitted to any other Institute or University for the award of any degree and are free from plagiarism.

Signature of Candidate: Name of Candidate: Soumendu Chakraborty Roll. No. 108151010

Certificate This is to certify that the above statements made by the candidate are correct to the best of my/our knowledge and belief.

Signature of Supervisor(s): Date: Name & Designation of Supervisor(s):Dr. Anand Singh Jalal Associate Professor

ii

Abstract The science of hiding secret information in another message is known as Steganography; hence the presence of secret information is concealed. It is the method of hiding cognitive content in same or another media to avoid recognition by the intruders. This thesis introduces new method wherein irreversible steganography is used to hide an image in the same medium so that the secret data is masked. The secret image is known as payload and the carrier is known as cover image. X-OR operation is used amongst mid level bit planes of carrier image and high level bit planes of data image to generate new low level bit planes of the stego image. Recovery process includes the X-ORing of low level bit planes and mid level bit planes of the stego image. Based on the result of the recovery, subsequent data image is generated. A RGB color image is used as carrier and the data image is a grayscale image of dimensions less than or equal to the dimensions of the carrier image. The proposed method greatly increases the embedding capacity without significantly decreasing the PSNR value.

iv

Acknowledgments I would like to express my gratitude to all of those who provided me with the resources and guidance to complete my thesis. I am highly indebted to my supervisor Dr. Anand Singh Jalal for his guidance and constant supervision as well as for providing necessary information regarding the thesis & also for his support in completing the thesis. His instrumental contribution in the proposed work has made it conceivable. I express my gratitude to Professor Charul Bhatnagar . Without her continuous guidance it would not have been possible to comprehend the concepts of digital image processing visually and experimentally. I sincerely thank Chancellor and Vice-chancellor of GLA University for providing a research oriented infrastructure. I would like to express my special gratitude and thanks to the Head of Department, Computer Engineering and Applications, Prof. Krishna Kant for giving me attention and time as and when needed. I would like to thank the all faculty members of the Department for their valuable suggestions. I sincerely thank Dr. Manas Kumar Mishra who made me realize the potential of my work and motivated me to work even harder. Special thanks to my colleague Suresh C. Raikwar who supported me at every step of my work. I would like to express my gratitude towards my parents and my wife for their kind cooperation and encouragement which helped me in completion of this thesis.

SOUMENDU CHAKRABORTY GLA University, Mathura

v

Contents Certificate

ii

Dedication

iii

Abstract

iv

Acknowledgments

v

List of Figures

viii

List of Tables

ix

1. Introduction

1

1.1 Overview and Motivation …………………………………....... 1.2 Issues In Image Steganography………………………………... ………………………………… 1.2.1 Steganography Security Measure………………………...

1 4

1.2.2 Steganalysis………………………....................................

5

1.3 Objectives ……………………………………………………...

6

1.4 Contributions of the Thesis ……………………………………

6

1.5 Outline of Thesis ………………………………………………

6

1.6 Summary ………………………………………………………

7

2. Literature Review

4

8

2.1 Introduction ……………………………………………………

8

2.2 State of Art Image Steganography..............................................

9

2.2.1 Irreversible Image Steganography……………………….

9

2.2.2 Reversible Image Steganography………………………..

11

2.3 Summary… ……………………………………………………

16

3. Proposed Methodology

17

3.1 Introduction ……………………………………………………

17

3.2 Proposed Method ………………………………………………

17

3.2.1 First Phase…….………………………………………….

18

3.2.2 Second Phase……..………………………………………

19

vi

Contents

3.2.3 Third Phase……………………………………………….

20

3.3 Summary ………………………………………………………

21

4. Results and Discussion

22

4.1 Data Set ………………………………… …………………….

22

4.2 Experiments and Results………………….……………………

23

4.2.1 Qualitative Analysis……………………………………...

24

4.2.2 Quantitative Analysis………………………………….…

26

4.3 Summary…..…………………...………………………………

30

5. Conclusions and Future Directions

31

5.1 Summary and Contributions …………………………………...

31

5.2 Proposed Method in Nutshell……………………………...…...

32

5.3 Future Work …………………………………………………...

32

References

34

List of Publications

37

vii

List of Figures

1.1

Steganographic Flow……………………………….…………….

2

3.1

Bit plane X-ORing algorithm…………………………………….

18

4.1

Test images (a)-(f) and Payload (g). (a) Lena; (b) Barbara; (c) Greens; (d) Baboon; (e) Boat; (f) Pepper; (g) Cameraman………..

22

4.2

Cover images (a)-(f) and stego images (g)-(l)…………………….

26

4.3

Comparison results (Lin et al., 2004), (Lee et al., 2008), and the proposed scheme ………………………………………………… Comparison of PSNR for different cover images and different payload using proposed algorithm…………………………………...

4.4

viii

28 29

List of Tables

4.1 Mean Square Error for different components of the cover image and Payload………………………………………………………………

27

4.2 Comparison results (Lin et al., 2004), (Lee et al., 2008), and proposed scheme ………………………………………………………...

27

4.3 Comparison of PSNR for different cover images and different size payload using proposed scheme …………………………………….

27

4.4 Comparison of Embedding Capacity and PSNR for state of art and the proposed scheme…………………………………………………

28

ix

Chapter 1 Introduction

1.1

Overview and Motivation

Modern day’s communication requires high level of security in transmission. There are two ways of achieving this: one by securing the channel and the other is by securing the message. Steganography is a well known and widely used technique that manipulates information (messages) in order to hide their existence. This technique has wide application area in computer science and image processing: It is used to secure messages of armed forces, corporate data, personal files, etc. Steganography is technique used in secure communication which hides the secret information [1]. Steganography hides secret message in a carrier file of some format and prevents the detection of the presence of the secret message. An intruder may intercept a carrier file containing the secret data without knowing the presence of secret data. Steganography, attempts to prevent an intruder from detecting the presence of secret information [2]. The objective of image steganography is to prevent the detection of the presence of secret data in the carrier image. On the other hand, steganalysis is to detect hidden in some media using steganography. In other words steganalysis try to falter steganography. Steganalysis is based on the principle that any modification in the carrier media may leave behind some kind of unusual bit pattern or some kind of debasement in the carrier which could be analyzed. Hence it is necessary that the steganography system makes absolutely certain that the hidden secret data is not detectable [3; 4; 5]. Steganography includes the hiding of various media like text, image, audio, video files in another media of same type or of different type, before the message hidden in the selected

Chapter 1

Introduction

media is transmitted to recipient. At the receiver’s end, reverse procedure is implemented to recover the original information [6].

Cover Image

Payload

Stego System Encoder

Payload

Stego System Decoder

Stego Image

Figure 1.1: Steganographic Flow

The general process of steganography is shown in Figure 1.1. Two fundamental expectations that a steganography algorithm should live up to are: (1) large number of secret bits can be embedded within the host image, so the algorithm must efficiently implement it, (2) the visual noise introduced due to embedding of the secret data should be at minimum level. Imperceptible stego-image quality is the most important feature of any steganography algorithm. There are several techniques that can be used to hide secret information inside an image like LSB Method, injection, substitution, and generation [6]. Two extensive approaches used in image steganography are

reversible image steganography

and irreversible image

steganography. Message to be embedded within an image is known as the payload. After embedding the payload in image (more specifically cover image) the resulting image called the stego image is sent to the authorized recipient, where the payload is recovered. Reversible image steganography derives its name from the mode of recovery of the payload wherein the recovered cover image is noiseless. On the contrary irreversible image steganography strives to achieve high capacity embedding without giving much emphasis on the carrier recovered during the extraction process. Reversible image steganography facilitates the easy detection of any alteration in the stego image whereas irreversible image steganography achieve high embedding capacity. Motivation for modern day’s image steganography techniques comes from the fact that no existing method is self-sufficient. Increasing the embedded information could cause easy detection by an attacker, whereas enhancing the security

Dept. of CEA, GLAU, Mathura

2

Chapter 1

Introduction

could increase the computation and communication overhead due to the fact that one secret message will take several transmissions. A trade-off between embedding capacity and the level of security needs a strong base for a security measure. The proposed scheme tries to achieve greater level of security with minimum computation time. The failure to cease the suspicion of any hidden data in an image is the reason for breakdown of any steganography system. The fundamental requirement for any secure steganography scheme is the ability to cover the hidden message in the cover image. A system is considered to be secure if a snooper cannot distinguish between cover image and the stego image. There are different measures for steganographic security. The most common measure is called detectability of a stego system. Detectability is defined as the relative entropy between the probability distribution of cover image and the stego image. Any steganography system is called  -secure if the relative entropy of the system is at most  [7]. A steganography scheme is said to be perfectly secure if detectability is zero. Reduction in detectability means reduced embedding capacity. Any image steganography scheme should optimize the embedding capacity to achieve minimum possible detectability taking into account the computational overhead. The maximum number of bits that can be embedded in a cover image and recovered from the stego image without violating the undetectability constraints is known as the steganographic capacity. The maximum steganographic capacity that an existent reversible steganographic scheme can achieve is approximately 3 bpp [8].

Perceptual consistency and robustness are the most

desirable attributes of any steganographic system. The capacity of the steganographic system is comprehensively enhanced by irreversible models of image steganography. Peak signal to noise ratio can be used as a measure of steganographic security where the embedding capacity of a model is outsized. Increasing secret information embedding capacity would mean straightforward steganalysis and detection of the hidden information. State of art steganalysis attempt to overpower any steganography scheme. Biggest challenge of a steganography scheme is to outsmart all steganalysis schemes [7]. The science of detecting the concealed message in a cover image is steganalysis. The battle between steganography and steganalysis is getting on since the evolution of the science of steganography. There are several ways in which steganalysis can wreck the structure of steganography. The most common methods

Dept. of CEA, GLAU, Mathura

3

Chapter 1

Introduction

are inspection of the inner structure of LSBs, Histogram analysis, feature vector analysis et cetera [7]. Primary goal of any image steganography scheme is to achieve high level of security with high capacity embedding, reduced noise, and minimum computation time.

1.2

Issues in steganography

Any steganography algorithm should consider two fundamental issues steganographic security measure and steganalysis. Steganographic security measure that any steganographic system incorporates has to be well defined. The system should satisfy any specific criteria applicable for steganographic security. Steganalysis deals with various analysis techniques employed on any algorithm to categorize the vulnerabilities associated with the algorithm. 1.2.1 Steganographic security measure There are different steganographic security measures as specified in [9; 10]. Any steganography method is said to be susceptible to attacks if cover and the srego image are distinctive. Cachin et al. [9] proposed that a steganographic algorithm is said to be

 -secure (   0) if the relative entropy between the cover image and the stego image probability distributions (PC and PS, respectively) is at most  . PC (1.1)  PS Equation (1.1) defines detectability D(.). Detectability increases with the increasing D( PC || PS )   PC  log

ratio PC

PS

. Increase in detectability means probability of detection of hidden message

using a steganalysis scheme is more. A steganographic technique is said to be perfectly secure if  = 0 (i.e. PC = PS). In other words the probability distribution of the cover and the stego images are same. There exists some steganography schemes which are perfectly secure. In our proposed algorithm, we intend to achieve perfectly secure steganography with higher embedding capacity. The detectability function is more suitable for analyzing image steganography schemes where the embedding capacity is very low. More appropriate measure for visual distortion in image steganography with high embedding capacity is Peak Signal to Noise Ratio (PSNR).

Dept. of CEA, GLAU, Mathura

4

Chapter 1

Introduction

PSNR is an objective measure which can be used to evaluate of extent of similarity between original image and the stego image [11; 12; 13]. PSNR is defined as; 2  I max  PSNR  10  log10   (dB)  MSE  where Imax = 255, maximum gray level for any grayscale image.

(1.2)

Mean squared error MSE (Moon et al., 2007) is defined to be

MSE 

1 M N (| C (i, j )  S (i, j ) |) 2  MN i 1 j 1

(1.3)

where M and N represent the number of pixels in a row and number of pixels in a column respectively. 1.2.2 Steganalysis There are two major approaches in steganalysis; the first approach is to develop a steganalysis scheme specific to a steganography scheme. The second approach is to develop a generalized steganalysis scheme which is applicable to most of the existing steganography schemes. Both approaches have their own advantages and disadvantages. A steganalysis scheme specific to a steganography scheme will detect the presence of hidden data accurately where as it may fail to detect the presence of hidden information in stego image generated using some other steganography scheme. On the right, a generalized steganalysis scheme may show optimum detection capability against almost all the existing steganography schemes and eventually completely fail to detect the hidden information embedded using some new steganography scheme. Steganalysis algorithms are said to be effective if they can detect the presence of hidden message accurately. Steganalysis schemes do not need to extract the hidden information the detection alone makes the schemes effective. The un authorized extraction of the secret data may be very difficult if the secret data is encrypted using a powerful encryption algorithm. However, there exists some steganalysis schemes proposed in recent literatures which effectively detect as well as estimates the size of the hidden secret information accurately.

Dept. of CEA, GLAU, Mathura

5

Chapter 1

1.3

Introduction

Objective

Image steganography is used to share the secret information by hiding the information into a cover image. Embedding capacity of any steganography scheme is measured as the number of bits that can be embedded into single pixel information of the cover. The major challenge in image steganography is to achieve high embedding capacity without violating the visual distortion constraint. Here we analyze different existing image steganography schemes and propose a new image steganography scheme with high embedding capacity.

1.4

Contributions of the Thesis

This thesis evaluates and analyzes the existing reversible and irreversible image steganography schemes and elaborates a new scheme with high embedding capacity. Image steganography methods hide secret information into some media so that the existence of secret message is concealed. One of the necessary attribute of any image steganography scheme is the hiding capacity. The amount of secret information that can be hidden into a cover image is called the hiding capacity. The scheme proposed in this thesis achieves high embedding capacity in the order of 8 bit per pixel. The problem that we face while increasing the embedding capacity is the level of distortion. Level of distortion introduced into the stego image obtained after embedding the secret information is directly proportional to embedding capacity. The proposed scheme maintains the level of distortion to a minimum so that the distortion introduced is unrecognizable by human visual system.

1.5

Outline of Thesis

This thesis introduces the concept of image steganography. Existing methods are categorized and elaborated. Rigorous analysis of existing scheme is done on the basis of known parameters. We also propose and analyze an efficient image steganography scheme in this thesis. The chapters are organized in this thesis as follows. Chapter 1 introduces the image steganography. Different categories of image steganography and their purpose are discussed in Chapter 1.

Dept. of CEA, GLAU, Mathura

6

Chapter 1

Introduction

Chapter 2 provides a detailed study of the existing image steganography schemes with proper analysis and evaluation. Chapter 3 introduces and elaborates the proposed image steganography scheme. Chapter 4 shows the results obtained based experiments conducted on using the scheme on relevant parameters. Analysis of the existing scheme on relevant parameters is done in Chapter 4. Chapter 5 provides a brief summary of the scheme and possibilities to improve the existing work.

1.6

Summary

Hiding secret information within an image is known as image steganography. Embedding capacity and the security measures are the two fundamental issues in image steganography. Achieving high embedding with higher level of security is the biggest challenge. In this thesis we propose an image steganography scheme with high embedding capacity. Proposed scheme maintains the security constraint with increasing embedding capacity.

Dept. of CEA, GLAU, Mathura

7

Chapter 2 Literature Review

2.1

Introduction

Extraction of hidden information from stego images is challenging task. Hence most of recent research work is going on around steganalysis. To get the better of steganalysis schemes many steganography schemes have been proposed, which explore the multiple dimensions of cover image. Many ideas and techniques have been proposed to secure data i.e., mainly concealing text or image in images. Steganography can be done in any domain namely spatial or frequency. The most fundamental image steganography method in frequency domain is spread spectrum image steganography (SSIS), which is image steganography method that uses a digital image as a cover. It is a data hiding scheme which hides large amount of data into a cover. SSIS is such a steganography scheme where original cover is immaterial in extraction of the secret data from the stego image. The authorized recipient only needs to have the secret key to recover the secret information without which the very presence of the secret data is undetectable. The most fundamental image steganography method in spatial domain is LSB. In Least Significant Bit replacement method the least significant bits of the cover image is replaced with the bits of the secret image. Different image formats supports these LSB steganography scheme. The embedding capacity of this steganography scheme is very high whereas the perceptibility is very low. As it is trivial enough to detect the presence of hidden information in LSB of the cover the extraction of the secret data is trivial as well.

Chapter 2

2.2

Literature Review

State of Art Image Steganography

Image steganography is used to hide secret information within an image [1]. Two major approaches used are reversible and irreversible image steganography. In reversible image steganography the cover image can be reconstructed accurately while extracting the payload from the stego image. The stego image is the image obtained after embedding the secret message in cover image. Most of the existing reversible image steganography schemes are very complex and achieve small embedding capacity. Embedding capacity can be increased by adaptive embedding of payload near sharper edges. More bits can be accommodated in sharper edges using adaptive selection. Irreversible image steganography schemes achieve higher embedding capacity with minimum computation time. Detection of hidden information in stego image resulting from irreversible stego system is straightforward. Many steganalytic schemes [2]-[4] have been proposed in literature, which can accurately detect the presence of secret information embedded using irreversible image steganography. These methods are prone to easy detection of the embedded information. Even though irreversible image steganography schemes achieve low computation time, low level of security degrades the performance of such system. Encryption of secret information could be one of the solutions. However, inclusion of encryption flawed the use of steganography as the fundamental need for image steganography is to eradicate the suspicion of hidden data. 2.2.1 Irreversible Image Steganography The simple method to conceal secret data is Least Significant Bit (LSB) replacement method. Secret message bits are embedded into the least significant bit plane of the cover image. However, it has its own limitations [14]. Steganalysis can be easily done on LSB replacement technique [15]. There are some existing steganalysis schemes which can be used to determine whether an image contains secret information, if the embedding process is as trivial as LSB. Cheddad et al. [7] proposed an adaptive steganographic approach that selects the specific region of interest (ROI) in the cover image. Data can be embedded in these regions. These regions are selected based on human skin tone detection. Based

Dept. of CEA, GLAU, Mathura

9

Chapter 2

Literature Review

on the size of data appropriate cover image should be selected so that secret information can be accommodated. It is very difficult to detect the presence of secret information in stego images is the embedding scheme used is of adaptive nature [16]. Tri way pixel value difference method proposed in [16] achieves high embedding capacity as well as high level of imperceptibility. Pixel value difference methods generally embed secret information using two pixel differences across one directional edge. The efficiency of these methods can further be increased using two pairs of two pixel value difference, which can be achieved by dividing the image into 2  2 blocks. As the fourth pair of pixel value consists of fourth pixel and the first pixel changing the difference value of fourth pair changes the difference values of first pair and the third pair, hence the fourth pair is discarded and three pairs are used to embed the secret data. The pre-procedure of the algorithm is to partition the cover image into non-overlapping 2  2 blocks with 4 pixels. The maximum PSNR that this scheme can attain is 38.89dB with approximate embedding capacity of 2 bpp. The result analysis shows that the proposed method achieves greater PSNR and better embedding capacity as compare to the scheme proposed by [16]. Babu et al. [17] proposed steganographic scheme which can be used to authenticate the secret information from the stego image itself. Recovery process verifies the integrity of the secret data by accurately recovering the original cover. In this method payload is transformed into frequency domain from the spatial domain using discrete wavelet transformation. The permutation of DWT coefficients are then embedded in the spatial domain of the cover image. This permutation is done with the verification code. DWT coefficients are used to generate the verification code. Thus the method can verify each row that has been modified by attacker. Moon et al. [18] proposed a fixed 4LSB method to embed an acceptable amount of data. It can easily be implemented and the degradation in the resulting image is not visually recognizable. In 4LSB method color bitmap images (24 bit and 8 bit i.e. 256 color palette images) are used as cover and wave files are used as the carrier media. Secret information in any format can be embedded into audio or image files using this method. After embedding the secret information the stego image or audio file can be sent through mail on the web. Authorized user with the appropriate password on receiving the mail can extract the secret message and decrypt the same. However, the fundamental drawback of this scheme is that the encoded message can

Dept. of CEA, GLAU, Mathura

10

Chapter 2

Literature Review

be easily recovered and even altered by third party. Lie et al. [19] proposed an adaptive method of variable length bit substitution instead of fixed length to adjust the hiding capacity. It is an adapted version of well known LSB method. The LSB of those pixel are used to embed the data whose grayscale values are distinct. Human visual sensitivity based piece wise mapping function is used so that the method can adapt according to the human visual system. The additional information required to decode the data is not very large. It is in the order of 3 bits only. Even though these methods [16; 17; 18; 19] increase the embedding capacity as well as level of security, the visual distortion introduced is a cause of concern. The proposed scheme enhances the embedding capacity while reduces the visual distortion introduced. Baekl et al. [11] proposed an image steganography code conversion is done to embed the secret data into an image. The characteristics of binary codes and gray codes are exploited using XOR operation to generate some meaningful patterns. This scheme requires that the N carrier images are shared amongst sender and receiver through a secure channel. Exclusive-OR operation is used to generate 2N meaningful binary patterns. In this method Bin2Dec function is used to compute the decimal equivalent of the given binary number. The result thus produced is denoted as X. X is generated by concatenating N bit binary number. The value of X is the deciding factor in deriving the results using an Exclusive OR operation among the 2N different N-bit binary codes and earlier results. Requirement of secure channel nullifies significance of steganography. So some methods are required so that the need of secured channel is nullified. The proposed scheme implements image steganography in such a way that the need for secure sharing of cover images is completely eliminated. 2.2.2 Reversible Image Steganography In [20] the author proposed a steganography method where variable length secret data is embedded in each block satisfying some constraint. This embedding scheme is lossless scheme as the cover can be recovered accurately. In this embedding scheme the original image is divided into a number of non overlapping blocks of m  n each. Then, sub images are selected in turn by scanning the image in horizontal and vertical directions for the embedding process. In each block, each element is defined as v0, v1, v2, . . . , vk−1 sequentially, and k represents the total number of pixels; in other words, total number of pixel k = m  n. In this scheme, an improved version of generalized

Dept. of CEA, GLAU, Mathura

11

Chapter 2

Literature Review

difference expansion is proposed which employ centralized difference expansion and achieves reduced distortion and high quality stego images. The size of payload for each block depends upon the complexity of the cover which reduces the image distortion and increases the hiding capacity. In order to the distortion some pixels in selected areas such as extreme edge areas are not used in embedding. All the blocks are classified into four types. Amount of data embedded in each block depends upon the type to which it belongs. Original image undergo severe distortion across edge areas in difference expansion methods, hence in [20] smoother areas are used to embed the secret information. The embedding capacity of this scheme is at most 1 bpp. There are many reversible image steganography schemes proposed in the literature which employ encryption to achieve higher level of security. Wu H. C. et al. proposed a reversible image steganography scheme [21], where the secret message is encrypted using either AES or DES. The encrypted bits are then embedded in a code tree computed from the frequency of absolute error values. Error values are computed using MED predictor [21]. The error values are used to build the hiding tree. This hiding tree is used hide the secret data. The range of each error value is [-255, 255]. The range of error values is adjusted to [0, 255] that is negative values are mapped to positive values. Since the negative error values do not have any significance in this scheme, the unsigned value of error is taken to generate absolute error value (AEV). AEV is the absolute error value set of X  Y absolute error values. In this scheme embedding process consists of four phases; predictive coding, encryption of secret data, building the hiding-tree, and secret bits embedding respectively. In the predictive coding phase, the host predictive image is generated from the original cover using MED predictor. H denotes the gray-scale cover image with X  Y pixels represented as C  ckl | 0  k  X , 0  l  Y , ckl  0,1, 2, , 255 . PD denotes the predictive

image

with

X Y

pixels

PD   pd kl | 0  k  X , 0  l  Y , pd kl  0,1, 2, , 255 .

represented The

error

values

as are

computed as difference between C and PD. These error values are used embed the secret data after encryption. E denotes the error value set of X  Y error values represented as ER  erkl | 0  k  X , 0  l  Y , erkl  0,1, 2, , 255 . The error value

Dept. of CEA, GLAU, Mathura

12

Chapter 2

Literature Review

erkl is computed as erkl  ckl  pd kl . In the proposed scheme to achieve the higher level of security the secret message is encrypted using state of art encryption methods such as AES or DES. ST denotes the secret data of M bits represented as ST  stv | 0  v  M , stv  0,1 where M < X  Y . ED denotes the encrypted secret

data. Encrypted data consists of M bits. Encrypted data is represented as ED  ed v | 0  v  M , ed v  0,1 . The encrypted data is computed using a mapping

function encr(.) as ED  encr ( ST , key ) where key is a secret key. Scheme proposed in [22] generates an intermediate image by converting a pair of pixel values of secret image into four hexadecimal values and then four hexadecimal values are converted to three decimal values. This intermediate image is then distributed and embedded into n cover images. To recover the secret image one has to gather all n stego images. The steganography scheme used in this method is straightforward and efficient. Detection of hidden information is so trivial that any steganalysis scheme can detect the presence of hidden information with more than 80% accuracy. A data hiding based on side-match vector quantization (SMVQ) has been proposed by Chang et al. [23]. Most of the research going on in reversible image steganography achieve high level of security and try to recover the original cover while extracting the payload. If the data is embedded after compression and the receiver wants to save compressed cover to save the space then the receiver has to extract the secret data from the cover. While extracting the secret data receiver has to recover the cover as well. The recovered payload has to be compressed to before storage in order to save the space. All these steps are cumbersome as well as time consuming. This reversible data hiding scheme is based on side match vector quantization (SMVQ) for digitally compressed images. In this scheme to store the compressed cover the receiver only has to extract the payload and can directly generate the SMVQ compression codes for the cover. For each block of cover image codeword is generated using SMVQ. These codeword are used to embed the secret data. If secret bit is equal to 0, the closest codeword generated by SMVQ is encoded. For a secret bit 1 the approximation of the first closest codeword and the second closest codeword is computed to replace the

Dept. of CEA, GLAU, Mathura

13

Chapter 2

Literature Review

closest codeword. Even though the proposed scheme effectively encodes the secret message, for a large payload the size of transformed index table can increase the space complexity of the steganography system. Moreover, the embedding capacity of the scheme is low compared to some other existing image steganography schemes. Reversible image steganography scheme proposed in [24] embeds secret information into cover image using histogram shifting. The pixel intensities with zero or minimum frequency are modified to embed the secret information. As each zero or minimum frequency pixel is modified by only one grayscale value the quality of the stego image is good. This reversible data hiding scheme hides secret data into the cover image in such a way that the original cover can be recovered without any distortion after extracting the secret data from the stego image. In this scheme minimum frequency points of the histogram are used to embed the secret data. Grayscale values of the pixels corresponding to minimum frequency points are slightly modified to embed the secret data. It can embed more data than many of the existing reversible data hiding algorithms. Peak signal to noise ratio (PSNR) of stego and corresponding cover image achieved by this method is at least 48dB. This lower bound of PSNR is much higher than the minimum required PSNR to effectively hide the secret data. The computational complexity of this technique is low and the execution time is short. However, embedding capacity of the scheme is very low. Hwang et al. [25] proposed a reversible image steganography scheme based on histogram shifting, which is an improvement over Ni et al. [24]. This scheme can effectively recover the original cover from the watermarked image while extracting the watermark. Watermarking methods cause serious degradation in the visual quality of the watermarked image. Using this algorithm the cover can be recovered without any distortion after the watermark is removed from the watermarked image. In this method maximum frequency points are used to embed the secret data. Location of the pixels having the maximum frequency in the histogram is stored. This location map is used to extract the secret data. The grayscale value corresponding to the peak points are slightly modified to embed the secret watermark. As the peak points of the histogram and the location map are enough to extract the secret information no additional information is required at the receiving end. The modification made to the pixels to embed the secret data is so negligible that high quality watermark images

Dept. of CEA, GLAU, Mathura

14

Chapter 2

Literature Review

can be produced. Recursive embedding can be used to increase the embedding capacity of this scheme as location map is generated to be used in extraction process. Lin et al. [26] proposed a block based reversible image steganography scheme, where the entire image is subdivided into blocks. Difference values are computed as the difference in pixel intensities of the first column and the rest of the columns. Secret information is embedded into these difference values. Instead of embedding data to peak points of the cover this scheme embeds data to the peak points of the difference image generated from the cover. Inverse transformation is applied to these peak points to generate the additional space required to embed the secret information. The hiding capacity of this scheme can be increased dynamically depending upon the application. Peak point in a histogram is a grayscale that the maximum numbers of pixels in that image have. In other words, large amount of secret data cannot be hidden in pixels having maximum frequency in an image. The characteristic of any image is such that the neighboring pixels have same grayscale value. Hence the difference image produces the difference values of neighboring pixels. The maximum frequency grayscale value in the difference image is zero. Hence to enhance the hiding capacity the histogram of the difference image produced from the original image is used to identify the peak points where the secret data is embedded. So the hiding capacity greatly increases if we consider the peak points of a difference histogram produced from the difference image of the corresponding original cover. Another block based image steganography scheme has been proposed in [27]. The center pixel in each block is used as the referential pixel. The difference values are computed from the referential pixel and the neighboring pixels in the block. Secret data is embedded in each block by modifying these difference values. The reversible hiding scheme proposed in [28] is an efficient method for watermarked images where the size of the secret data is small. The method computes an auxiliary image using feasible image interpolation [29]. The difference values of original and auxiliary image are used to hide the secret data. Kim et al. [30] proposed a method where cover image is sampled into a number of sub-images. One of the subimages is taken as the reference image. Difference values computed from the reference image and rest of the sub-images are used to embed the secret payload. Reversible image steganography schemes have very low embedding capacity. An alternate to reversible image steganography is irreversible image steganography. High

Dept. of CEA, GLAU, Mathura

15

Chapter 2

Literature Review

embedding capacity of irreversible image steganography draws researchers to work in this area. Security is big concern in irreversible image steganography. Easy detection of hidden data is possible with some powerful steganalytic tools.

2.3

Summary

There are two major approaches in image steganography; reversible and irreversible. In reversible image steganography the cover image can be accurately recovered from the stego image while extracting the hidden information. Reversible methods achieve low embedding capacity compared to irreversible methods. Reversible methods are more secure than their irreversible counterparts. Reversible as well as irreversible methods embed secret data in spatial as well as frequency domain.

Dept. of CEA, GLAU, Mathura

16

Chapter 3 Proposed Methodology

3.1

Introduction

The higher embedding capacity is the focal point of the proposed scheme. The proposed scheme also eliminates the need of the cover image to recover the payload. Payload can be directly recovered from the stego image. It reduces the need for a secure channel to share the cover images before the start of any secret communication. The proposed algorithm embeds secret image into different component of the RGB cover image. It achieves higher embedding as it exploits every part of available space in a cover. Figure 3.1 illustrates the operational flow of the proposed approach.

3.2

Proposed Method

The proposed algorithm has three phases. First phase includes bit plane slicing of the cover image as well as the payload into eight bit planes. This algorithm takes any RGB color image as a cover image, so RGB components are separated before the bit plane slicing. Second phase performs X-OR operation amongst some of the mid level bit planes of the cover image and high level bit planes of the payload and the result is injected into the remaining lower level bit planes of the cover image. The third phase recovers the payload from the stego image by X-ORing the mid level bit planes of the stego image with low level bit planes of the stego image. The resulting bit planes are the bit plane of the grayscale payload. First higher order two or three bit planes are considered as high level bit planes, next two or three higher bit planes are considered to be mid level bit planes.

Chapter 3

Proposed Methodology

A module Bit Plane Slicing extracts the bit planes of different component of the cover image and the bit planes of the payload. The module bit plane slicing has four sub modules R component bit plane slicing, G component bit plane slicing, B component bit plane slicing, and payload bit plane slicing. Cover Image

RGB extraction

R Component

Payload Bit Planes

G Component

B Component

Payload

Bit Plane Slicing

R bit Planes

G bit Planes

B bit Planes







Merge Bit Planes

Stego Image Figure 3.1: Bit plane X-ORing algorithm

3.2.1 First phase Different components of the RGB cover image are separated in this phase. The major task of bit plane extraction of cover as well as payload image is done for further processing. This phase is operational at transmission end of any communication. Let C be the RGB cover image. The R, G, and B components of the cover image C are separated. Let us denote these components as CR, CG and CB respectively. Each component is sliced into eight bit planes.

Dept. of CEA, GLAU, Mathura

18

Chapter 3

Proposed Methodology

C

C

R

C G C B   split _ comp  C 

(3.1)

(1)|C R (2)|C R  3 |.....| C R  7  | C R  8    dec2binp  CR , 8 

(3.2)

R

 C 1 | C  2  | C  3 |.....|C  7  |

CG  8    dec2binp  C G , 8 

(3.3)

 C 1 | C  2  | C 3 |.....| C  7  |

C B  8    dec2binp  C B , 8 

(3.4)

G

B

G

B

G

B

G

B

The function split_comp() separates three components of an RGB image. Equations (3.2)-(3.4) slices each component into eight bit planes, where CR(1), CG(1) and CB(1) are the highest level bit planes of CR, CG, and CB respectively and CR(8), CG(8) and CB(8)are the lowest level bit planes of CR, CG, and CB respectively. Similarly, payload P is sliced as in equation (3.5).

 P 1 |P  2  | P 3 | P  4  | P  5  | P  6  | P  7  | P 8  

 dec2binp  P, 8 

(3.5)

The function dec2binp() converts any plane in decimal into its binary equivalent planes and returns an array of planes of size as specified in the second parameter of the function. 3.2.2 Second Phase This phase performs Bit plane X-ORing on different bit planes of cover and payload images. First three bit planes of payload P; P(1), P(2), and P(3) are X-ORed with mid level three bit planes of CR and the result is stored in the low level three bit planes of CR. The basic idea behind X-ORing the mid level bit planes of cover image with the high level bit planes of the payload is that some of the characteristics of mid level bit planes

remain intact in the resulting bit planes due to the nature of the X-OR

operation. When we insert these planes in the lower level bit planes of the cover image, we are essentially transforming the bit patterns of the lower level bit planes into the bit patterns of the mid level bit planes of the cover image. This diminishes the expected distortion in the resulting stego image. C RM  6   XOR  C R  5  , P 1 

(3.6)

C RM  7   XOR  C R  4  , P  2  

(3.7)

C RM  8   XOR  C R  3 , P  3  

(3.8)

Dept. of CEA, GLAU, Mathura

19

Chapter 3

Proposed Methodology

Similarly, P(4), P(5), and P(6) are X-ORed with mid level three bit planes of CG and the result is stored in the low level three bit planes of CG. CGM  6   XOR  CG  5  , P  4  

(3.9)

CGM  7   XOR  CG  4  , P  5  

(3.10)

CGM  8   XOR  C G  3 , P  6  

(3.11)

Last two bit planes of the payload are X-ORed with mid level two bit planes of CB and the result is stored in the low level two bit planes of CB. C BM  7   XOR  C B  6  , P  7  

(3.12)

C BM  8   XOR  C B  5  , P  8  

(3.13)

Red, green and blue components of the stego image are reconstructed using the high level bit planes and modified mid level bit planes of each component of the cover image. C RM  bin2decp  C R 1 |C R  2  |C R  3 |.....| C RM  7  | C RM  8  

(3.14)

CGM  bin2decp  C G 1 |CG  2  | CG  3 |.....|C GM  7  | C GM  8  

(3.15)

C BM  bin2decp  C B 1 |C B  2  |C B  3 |.....| C BM  7  | C BM  8  

(3.16)

The function bin2decp() converts an array of binary planes into a decimal plane. Stego image S is formed by merging these three modified components.

S  Merge _ Comp  CRM , CGM , CBM 

(3.17)

Merge_Comp() merges three components to form a RGB image. So, S is the stego image obtained after phase three. 3.2.3 Third Phase This phase is the recovery phase. Payload is recovered from the stego image at the receiving end. Recovery process is just the reverse of phase two. Stego image is split into the equivalent red, green and blue component.

C

RM

C GM C BM   split _ comp  S 

(3.18)

The function split_comp() separates three components of an RGB image. Now, individual color components are sliced into bit planes.

Dept. of CEA, GLAU, Mathura

20

Chapter 3

Proposed Methodology

 C 1 | C  2  | .. | C  7  | C 8 

 dec2binp  C RM , 8 

(3.19)

 C 1 | C  2  | .. | C  7  | C 8 

 dec2binp  CGM , 8 

(3.20)

C BM  8    dec2binp  C BM , 8 

(3.21)

RM

GM

RM

RM

GM

GM

 C 1 | C  2  | .. | C  7  | BM

BM

BM

RM

GM

The payload bit planes are recovered by X-ORing the respective bit planes of each component. RP 1  XOR  C RM  5  , C RM  6  

(3.22)

RP  2   XOR  C RM  4  , C RM  7  

(3.23)

RP  3  XOR  C RM  3 , C RM  8  

(3.24)

RP  4   XOR  CGM  5  , C GM  6  

(3.25)

RP  5   XOR  CGM  4  , CGM  7  

(3.26)

RP  6   XOR  CGM  3 , CGM  8  

(3.27)

RP  7   XOR  C BM  6  , C BM  7  

(3.28)

RP  8   XOR  C BM  5  , C BM  8  

(3.29)

These recovered bit planes are combined to recover the payload. RP  bin2decp  RP 1 |RP  2  | RP  3  |.....|RP  7  | RP 8  

(3.30)

Equation (3.30) gives the recovered payload. This payload is obtained from the stego image and this method does not require any cover image to be shared.

3.3

Summary

The proposed scheme embeds secret image into different components of an RGB color image. Proposed scheme has three phases. Different components of the cover image are segregated. Each component is then divided into eight bit planes. Bit planes of each component are used to embed the bit planes of the secret image. Proposed scheme achieves high embedding capacity in the order of 8 bit per pixel. Distortion introduced into the resulting stego image is visually unrecognizable.

Dept. of CEA, GLAU, Mathura

21

Chapter 4 Results and Discussions

4.1

Data Set

To check the performance of the proposed method we embedded the same payload in six different cover images Lena, Barbara, Boat, Greens, Pepper, and Baboon of different complexity. All the cover images were selected from standard image set used in other state-of-art. art. The cover images and the the payload are shown in Figure 4.1(a)-(f) and Figure 4.1(g) (g) respectively. In each instance the size of the cover image (M  N )

and the size of the payload (m  n) satisfied the following constraints

mM &nN.

a

b

c

d

e

f

g

Figure 4.1:: Test images (a)-(f) (a) (f) and Payload (g). (a) Lena; (b) Barbara; (c) Greens; (d) Baboon; (e) Boat; (f) Pepper; (g) Cameraman.

Chapter 4

Results and Discussions

By taking a cover image of size slightly greater than the size of payload the noise in the sego image can be reduced further. The secret bit planes are produced by performing X-OR operation amongst the bit planes of the cover image and the bit planes of the payload. The overall payload is same and independent of the complexity of the cover image. With our scheme 8 bits of the secret data are inserted in each pixel of the cover image, meaning there by 8 bit pixel information of the payload is embedded in 24 bit pixel information of the cover image. The major problem in increasing the embedding capacity is the rise in visual distortion in the stego image. It is generally known that the distortion of the stego image is hard to detect by the human eyes as long as the PSNR value is greater than or equal to 30 dB (Lee et al., 2008). The amount of data inserted in each instance of the experiments is kept constant to analyze the distortion introduced in the stego image, while the cover image in each instance is changed. Even though, the distortion is visually unrecognizable the introduction of noise is inevitable.

4.2

Experiments and Results

The fundamental method used to determine the noise in stego image is peak-signal-tonoise-ratio (PSNR). Efficiency of any image steganography algorithm depends on hiding capacity and embedding efficiency. So, we consider both aspects to analyze the results. PSNR is an objective measure for subjective evaluation of degree of similarity between an original image and a stego image (Baekl et al., 2010). PSNR is defined as;

 I2  PSNR  10  log10  max  (dB)  MSE 

(4.1)

where Imax = 255, maximum gray level for any grayscale image, and the mean squared error MSE (Moon et al., 2007) is defined to be

MSE 

1 M N  (| C (i, j )  S (i, j) |)2 MN i 1 j 1

(4.2)

where M and N represent the number of horizontal and vertical pixels of the images respectively. We are using a RGB image as cover image, so the noise introduced in each component of the stego image has to be evaluated. The PSNR of

Dept. of CEA, GLAU, Mathura

23

Chapter 4

Results and Discussions

the stego image is defined as the average of the PSNR calculated for different components of the stego image. Let CR(i, j) be the pixel intensity of R component of the cover image and SR(i, j) be the pixel intensity of R component of the stego image. Similarly for green and blue components we have CG(i, j), CB(i, j) and SG(i, j),SB(i, j). MSE for R, G, and B components are calculated using following equations:

MSER 

1 M N (| CR (i, j )  S R (i, j ) |) 2  MN i 1 j 1

(4.3)

MSEG 

1 M N (| CG (i, j )  SG (i, j ) |) 2  MN i 1 j 1

(4.4)

MSEB 

1 M N (| CB (i, j )  S B (i, j ) |) 2  MN i 1 j 1

(4.5)

PSNR for R, G, and B components are calculated using following equations: 2  I max  PSNRR  10  log10   (dB)  MSER 

(4.6)

 I2 PSNRG  10  log10  max  MSEG

  (dB) 

(4.7)

 I2  PSNRB  10  log10  max  (dB)  MSEB 

(4.8)

The PSNR is the average of PSNRR, PSNRG, and PSNRB. PSNR 

PSNRR  PSNRG  PSNRB 3

(4.9)

4.2.1 Qualitative Analysis Figure 4.2(a)-(l) shows the cover and the resulting stego images. The image “Lena” is a cover image with high complexity even then we keep the size of payload same. Each instance of the experiment has the same payload (Cameraman). Even though eight pixel information of the payload is distributed over different bit planes of the cover image, the visual distortion is almost unrecognizable.

Dept. of CEA, GLAU, Mathura

24

Chapter 4

Results and Discussions

a

b

g

h

c

i

d

j

Dept. of CEA, GLAU, Mathura

25

Chapter 4

Results and Discussions

e

f

k

l

Figure 4.2: Cover images (a)-(f) and stego images (g)-(l) Change in cover image does not affect the visual features of the payload after extraction. Visual distortion in the stego image is minimal, if we change the cover image for the same payload. Complexity of the image “Barbara” is different in many respects even then the distortion in the stego image for the same payload (Cameraman) is negligible. The fundamental requirement of any image steganography algorithm is minimum visual distortion in the resulting stego image and it should be same for different cover images. The results illustrate that the proposed algorithm conforms to these requirements. 4.2.1 Quantitative Analysis Mean Square Error (MSE) of different components of cover images are shown in Table 4.1. Table 4.1 illustrates that the MSE of blue component is greatest in most of the cases. So in the proposed scheme significant noise reduction in stego image has been achieved by embedding minimum number of bit planes in blue component of the cover image.

Dept. of CEA, GLAU, Mathura

26

Chapter 4

Results and Discussions

Table 4.1 Mean Square Error for different components of the cover image and Payload Mean Square Error (decimal)

Images

Red

Green

Blue

Barbara

8970

8190

9330

Lena

6070

6130

7210

Green

10100

9480

11300

Baboon

8350

6420

8740

Boat

6164

6164

6164

Pepper

6522

10525

9092

Table 4.2 Comparison results (Lin et al., 2004), (Lee et al., 2008), and proposed scheme PSNR(dB) (Lin et al., 2004)

PSNR(dB) (Lee et al., 2008)

Barbara

38.12

34.74

Average PSNR(dB) Proposed Scheme 40.08

Lena

38.52

34.32

39.94

38.38

34.27

40

38.26

34.84

40.04

Boat

38.40

34.41

40.07

Pepper

38.45

34.24

40.04

Images

Greens Baboon

Embedded Data

4,60,800 bits

Table 4.3 Comparison of PSNR for different cover images and different size payload using proposed scheme Payload (Bits)

Barbara PSNR(dB)

Lena PSNR(dB)

Greens PSNR(dB)

Baboon PSNR(dB)

Boat PSNR(dB)

Pepper PSNR(dB)

20,000

40.05

39.97

39.94

39.89

40.07

40.05

80,000

40.09

39.95

39.93

40

39.99

40.02

3,20,000

40.05

39.96

39.99

40.04

40.01

40.06

4,60,800

40.08

39.94

40

40.04

40.07

40.04

Dept. of CEA, GLAU, Mathura

27

Chapter 4

Results and Discussions

Table 4.4 Comparison of Embedding Capacity and PSNR for state of art and the proposed scheme Steganography Schemes (Luo et al., 2010) (Moon et al., 2007) Proposed

Maximum Embedding Capacity (bpp) 0.5 4 8

Maximum PSNR (dB) 54.1 34.3 40.6

Figure 4.3: Comparison results (Lin et al., 2004), (Lee et al., 2008), and the proposed scheme

Dept. of CEA, GLAU, Mathura

28

Chapter 4

Results and Discussions

Figure 4.4: Comparison of PSNR for different cover images and different payload using proposed algorithm.

Results shown in Table 4.2 illustrate comparison of proposed scheme with (Lin et al., 2004) and (Lee et al., 2008). The proposed scheme achieves approximately same PSNR (dB) for different cover images, for the same payload. Hence, the choice of cover image is not a critical consideration for the proposed algorithm. A single image is enough to carry the entire payload and one need not consider a series of images for data embedding. For different sized payload, the change in PSNR for different cover images is shown in Table 4.3. Results shown in Table 4.3 clearly illustrate the fact that for a given cover image the PSNR is stable for different sized payloads. Maximum embedding capacity and the maximum PSNR achieved in state of art and proposed scheme has been depicted in Table 4.4. The proposed method attains higher embedding capacity while maintaining the desired PSNR. The proposed algorithm is efficient enough in attaining higher and stable PSNR as compared to (Lin et al., 2004) and (Lee et al., 2008) as shown in Figure 4.3.The change in PSNR for different payload is shown in Figure 4.4. For increasing payload the proposed algorithm confer minimum change in PSNR. Even if the size of the payload is increased, the PSNR value never goes below the minimum required value of 30dB.

Dept. of CEA, GLAU, Mathura

29

Chapter 4

Results and Discussions

4.3 Summary Experimental results show that the distortion introduced into the stego image is visually unrecognizable. Suitable quantitative measure for distortion, if the embedding capacity is very high, is PSNR. PSNR for visually undistorted stego image is at least 30 dB. Proposed scheme achieves much higher PSNR value in the order of 40 dB at any embedding rate. Even though we increase the size of the payload proposed scheme produces high quality stego image.

Dept. of CEA, GLAU, Mathura

30

Chapter 5 Conclusion and Future Direction

5.1

Summary and Contribution

In this thesis we propose an efficient irreversible image steganography scheme. This scheme is efficient in terms of embedding capacity. This thesis presents a brief overview of image steganography. Recent reversible as well as irreversible image steganography schemes reported in the literature are elaborated and evaluated in this thesis. Chapter 1 provides an introduction to general image steganography process. General terms such as cover or carrier image, payload and stego image are defined. We present a detailed description of each sub process constituting the general flow of image steganography. Some of the fundamental issues related to image steganography have been described in this chapter. The motivation to work on the irreversible image steganography follows from the challenges that exist in this area. Chapter 2 categorizes the existing image steganography schemes. Broadly we can categorize the image steganography as reversible and irreversible image steganography. Detailed account of each of the state of art reversible as well as irreversible image steganography schemes has been presented in this chapter. From early age of image steganography major developments as well as the gradual growth have been depicted visibly. Chapter 3 identifies the possibilities to improve upon the existing schemes and proposes an efficient image steganography scheme which achieves higher embedding capacity. The detailed clarification as to why and how the proposed scheme should work has been presented in this chapter. The proposed scheme has been depicted as well as formulated in the most logical manner.

Chapter 5

Conclusion and Future Direction

Chapter 4 presents detailed evaluation parameters and analysis of the proposed scheme. The proposed scheme is evaluated using qualitative analysis where quality of the stego image is shown and compared with the quality of the original cover. The proposed method is compared with all most all the state of art image steganography schemes presented in chapter 2. Detailed comparative study of the proposed scheme and some the existing methods are also shown in this chapter. Chapter 5 summarizes the thesis. Brief summary of all the chapters is given. The proposed methodology is summarized. A brief overview of the future possibility is also provided in this chapter.

5.2

Proposed Method in Nutshell

The proposed algorithm exploits the multiple dimensions of a RGB image to embed a secret image. The method not only achieves high bit per pixel (bpp) embedding but also manages reduced MSE in the resulting stego-image. Bit Plane X-ORing is used to modify the bit planes of the different components of the cover image. The resulting stego image is of the same size as the size of the cover image, which nullifies the detection of hidden information in the carrier. Recovery involves X-ORing of certain bit plains of the stego image. The proposed method attains higher PSNR and embedding capacity. In this scheme the need for shared cover images for recovery of the payload is taken out of the picture. The characteristic of X-OR operation is such that it can be used to embed the bits of the secret image into low level bit planes of the cover without significant increase in distortion. As bit planes of the secret image are X-ORed with mid level bit planes of the cover image and resulting bit planes are embedded into low level bit planes of the cover image some of the characteristics of mid level bit planes of the cover are reproduced in the low level bit planes of the cover. As a result the distortion level does not go out of proportion with increased embedding.

5.3

Future Work

Further improvements are possible on this method if any other correlation could be found amongst the R, G, and B components of the cover-image and the payload. Our

Dept. of CEA, GLAU, Mathura

32

Chapter 5

Conclusion and Future Direction

future work will involve working with color images as payload. A method also needs to be developed for grayscale payload images whose pixel depth is more than 8 pixels, as is the case in technical uses like medical imaging or remote sensing.

Dept. of CEA, GLAU, Mathura

33

References 1.

Clair and Bryan, “Steganography: How to Send a Secret Message”,www.strangehorizons.com/2001/20011008/steganography.s html, 2001.

2.

A. Westfeld, and G. Wolf, “Steganography in a Video conferencing system”, in proceedings of the second international workshop on information hiding, LNCS Vol. 1525, Springer, pp. 32-47, 1998.

3.

C. C. Lin, and W. H. Tsai, “Secret Image Sharing with Steganography and Authentication”, Journal of Systems and Software, Vol. 73, No. 3. pp. 405-414, 2004.

4.

K. Rabah, “Steganography - The Art of Hiding Data”, Information technology Journal, Vol. 3, No. 3, 2004.

5.

T. Morkel, J. H. P. Eloff and M. S. Olivier, “An Overview of Image Steganography”, in Proceedings of the Fifth Annual Information Security South Africa Conference (ISSA2005), Department of Computer Science, University of Pretoria, SA, 2005.

6.

R. Krenn, “Steganography and http://www.Krenn.nl/univ/cry/steg/article.pdf, 2004.

7.

A. Cheddad, J. Condell, K. Curran, P. McKevitt “Digital image Steganography: Survey and analysis of current methods”, Signal Processing , Elsevier, Vol. 90 , pp.727–752, 2010.

8.

Y.H. Yu, C.C. Chang, Y.C. Hu “Hiding secret data in images via predictive coding”, Pattern Recognition, Vol. 38, pp. 691–705, 2005.

9.

C. Cachin, “An information-theoretic model for steganography”, 2nd International Workshop Information Hiding, LNCS Vol. 1525, pp. 306–318, 1998.

10.

J. Zollner, H. Federrath, H. Klimant, A. Pfitzman, R. Piotraschke, A. Westfeld, G. Wicke and G. Wolf, “Modeling the security of steganographic systems”, 2nd Information Hiding Workshop, LNCS Vol. 1525, pp. 345–355, 1998.

11.

J. Baekl, C. Kim, P. S. Fisherl and H. Cha, “(N, 1) Secret Sharing Approach Based on Steganography with Gray Digital Images”. In proceedings of IEEE International Conference on Wireless Communications, Networking and Information Security (WCNIS), pp. 325 – 329, 2010.

12.

B. Li, J. He, J. Huang and Y. Q. Shi “A Survey on Image Steganography and Steganalysis”, Journal of Information Hiding and Multimedia Signal Processing, Vol. 2, No. 2, pp. 142–171, 2011.

Steganalysis”,

References 13.

C.Y. Yang, C.H. Lin and W.C. Hu “Reversible Data Hiding for HighQuality Images Based on Integer Wavelet Transform”, Journal of Information Hiding and Multimedia Signal Processing, Vol. 3, No. 2, pp. 142–150, 2012.

14.

R.J. Anderson, and F. A. P. Petitcolas, “On the limits of the Steganography”, IEEE Journal Selected Areas in Communications, Vol. 16, No. 4, pp. 474-481, 2001.

15.

S. Dumitrescu, W.X. Wu and N. Memon, “On steganalysis of random LSB embedding in continuous-tone images”, in Proceedings of International Conference on Image Processing, Rochester, NY, pp. 641-644, 2002.

16.

K. C. Chang, C. P. Chang, P. S. Huang and T. M. Tu, “A novel image steganographic method using Tri-way pixel value Differencing”, Journal of multimedia, Vol. 3, No. 2, 2008.

17.

K. S. Babu, K. B. Raja, K. K. Kumar, T. H. Manjula Devi, K. R. Venugopal, and L. M. Pataki, “Authentication of secret information in image steganography”, IEEE Region 10 Conference TENCON, pp. 1-6, 2008.

18.

S. K. Moon and R.S. Kawitkar, “Data Security using Data Hiding”, IEEE International conference on computational intelligence and multimedia applications, Vol. 4, pp. 247- 251, 2007.

19.

W. N. Lie and L. C. Chang, “Data Hiding in images with adaptive numbers of least significant bits based on human visual system”, IEEE international conference on image processing, Vol. 1, pp. 286-290, 1999.

20.

C. C. Lee, H. C. Wu, C. S. Tsai and Y. P. Chu, “Adaptive lossless steganographic scheme with centralized difference expansion”, Pattern Recognition, Elsevier, Vol. 41 , pp. 2097 – 2106, 2008.

21.

H. C. Wu, H. C. Wang, C. S. Tsai and C. M. Wang, “Reversible image steganographic scheme via predictive coding”, Displays, Elsevier, Vol. 31, pp. 35–43, 2010.

22.

G. Ulutas, M. Ulutas and V.V. Nabiyev, “Secret image sharing with reversible capabilities”, International Journal Internet Technology and Secured Transactions, Vol. 4, No. 1, pp. 1-11, 2012.

23.

C.C. Chang, W.L. Tai, and C.C. Lin, “A reversible data hiding scheme based on side- match vector quantization”, IEEE Transactions on Circuits and Systems for Video Technology, Vol. 16, No. 10, pp. 1301–1308, 2006.

24.

Z. Ni, Y.Q. Shi, N. Ansari, W. Su, “Reversible data hiding”, IEEE Transactions on Circuits and Systems for Video Technology, Vol.16, No. 3, pp. 354–362, 2006.

Dept. of CEA, GLAU, Mathura

35

References 25.

J. Hwang, J.W. Kim, J.U. Choi, “A reversible watermarking based on histogram shifting”, LNCS Vol. 4283, pp.348–361, 2006.

26.

C.C. Lin, W.L. Tai, C.C. Chang, “Multilevel reversible data hiding based on histogram modification of difference images”, Pattern Recognition, Vol. 41, No. 12, pp.3582–3591, 2008.

27.

P.Y. Tsai, Y.C. Hu, H.L. Yeh, “Reversible image hiding scheme using predictive coding and histogram shifting”, Signal Processing, Vol. 89, No. 6, pp.1129–1143, 2009.

28.

L. Luo, Z. Chen, M. Chen, X. Zeng, Z. Xiong, “Reversible image watermarking using interpolation technique”, IEEE Transactions on Information Forensics and Security, Vol. 5, No. 1, pp.187–193, 2010.

29.

L. Zhang, X. Wu, “An edge-guided image interpolation algorithm via directional filtering and data fusion”, IEEE Transactions on Image Processing, Vol. 15, No. 8, pp.2226–2238, 2006.

30.

K. Kim, M. Lee, H. Lee, H. Lee, “Reversible data hiding exploiting spatial correlation between sub-sampled images”, Pattern Recognition, Vol. 42, No. 11, pp.3083–3096, 2009.

Dept. of CEA, GLAU, Mathura

36

List of Publication 1. Soumendu Chakraborty, Anand Singh Jalal and Charul Bhatnagar, ‘An efficient bit plane X-OR algorithm for irreversible image steganography’, International Journal of Trust Management in computing and communications (IJTMCC), Inderscience, Vol. 1, No. 2, pp. 140-155, 2013.

Dept. of CEA, GLAU, Mathura

37