International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________394IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________More Secured Steganography Model with High Concealing Capacity by usingGenetic Algorithm, Integer Wavelet Transform and OPAPJyoti 1M.Tech. Scholar, Digital Communication, Rajasthan Technical University-KotaDepartment of Electronics & Communication EngineeringSobhasaria Engineering College, Sikar, Rajasthan, IndiaE-mail: email@example.comMd. Sabir 2Assistant ProfessorDepartment of Electronics & Communication EngineeringSobhasaria Engineering College, Sikar, Rajasthan, IndiaE-mail: firstname.lastname@example.orgAbstract: Steganography is an art of writing for conveying message inside another media in a secret way that can only be detectedby its intended recipient. There are security agents who would like to fight these data hiding systems by steganalysis, i.e.discovering covered secret messages and rendering them useless. Steganalysis is the art of detecting the messages existence,message length or place of message where it is to be hidden in covered media and blockading the covert communication. There iscurrently no more secured steganography system which can resist all steganalysis attacks such as visual attack, statistical attack(active and passive) or structural attack. The most notable steganalysis algorithm is the Reversible Statistical attack which detectsthe embedded message by the statistic analysis of pixel values. To maintain the security against the Reversible Statistical analysis,the proposed work presents a new steganography model based on Genetic Algorithm using Integer Wavelet Transform. Wepresent a novel approach to resolve such problems of substitution technique of image steganography. Using the proposed GeneticAlgorithm and Reversible Statistical analysis Algorithm, the system is more secured against attacks and increases robustness. Therobustness would be increased against those attacks which try to reveal the hidden message and also some unintentional attackslike noise addition as well. In this proposed work, we studied the steganographic paradigm of data hiding in standard digitalimages. In recent literature, some algorithms have been proposed where marginal statistics are preserved for achieving morecapacity and more security. This proposed system presents a novel technique to increase the data hiding capacity and theimperceptibility of the image after embedding the secret message. In proposed work Optimal Pixel Adjustment Process alsoapplied to minimize the error difference between the cover and stego image. By this work best results have been obtained ascompared to existing works. The proposed steganography model reduces the embedding error and provides higher embeddingcapacity. Detection of message existence will be very hard for those stego images that produced using the proposed method. Thiswork shows the highest embedding capacity and security against Reversible Statistical attack.Keywords: Genetic Algorithm, IWT, OPAP, RS Analysis._____________________________________________________*****______________________________________________________I. INTRODUCTIONThe standard and thought of “What You See Is What YouGet (WYSIWYG)” which we have a tendency toencounter typically while printing images or othermaterials, is no longer precise and would not mislead asteganographer as it does not always hold true. Images areover what we see with our Human Visual System (HVS);therefore, they can convey over 1000 words .Steganography, the art of hiding messages inside othermessages, is now gaining more popularity and is beingused on various media such as text, images, sound, andsignals. However, none of the existing schemes can yetdefend against all type of detection attacks. Using GA‟sthat are based on the procedures of natural genetics andthe theory of evolution, we can design a general methodto guide the steganography process to the best position fordata hiding .Steganography is the art of hiding informationimperceptibly in a cover media. The word"Steganography" is Greek word which means “concealedwriting”. Where Stegano means "protected or covered”and graphy - “to write". Steganography is the art andscience of hiding communication; a steganographicsystem so embeds hidden content in unremarkable covermedia so as not to arouse an eavesdropper‟s suspicion. Inthe past, individuals used hidden tattoos or invisible ink toconvey steganographic content. Today, personal computer(PC) and network technologies give easy-to-usecommunication channels for steganography.Essentially, the information-hiding process in asteganographic system starts by identifying a covermedium‟s redundant bits (those that can be modifiedwithout destroying that medium‟s integrity) . The
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________395IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________embedding process creates a stego medium by replacingthese redundant bits with data from the hidden message.Modern steganography‟s goal is to stay its mere presenceundetectable, but steganographic systems, thanks to theirinvasive nature, leave behind detectable traces within thecover medium. Although secret content is not discovered,the very existence of it is: modifying the cover mediumchanges its statistical properties, thus eavesdroppers cannotice the distortions within the resulting stego medium‟sstatistical properties. The strategy of finding thesedistortions is named statistical steganalysis.The purpose of steganography is to hide the presence ofcommunication while the purpose of cryptography is tomake the communication incomprehensible by modifyingthe bit streams using secret keys. The advantage ofsteganography, over cryptography is that the attackers arenot attracted towards communicating messages betweensender and receiver while the encrypted messages attractthe attackers. Steganalysis is a method of detecting themessage hidden in a cover media and to extract it.Changes will be apparent in the statistical property ofimage if the secret message bits are inserted in image. Thestrength of the steganography is measured bysteganalysis. RS steganalysis is one of the most reliablesteganalysis which performs statistical analysis of thepixels to successfully detect the message hidden in theimage. However, steganography method to detect thepresence of secret message by RS attack/analysis isdifficult in case of color images. Retention of visualquality of the image is also imperative. It is worth to notethat genetic algorithm optimizes security and also thequality of the image. It belongs to class of evolutionaryalgorithms, which imitates the process of naturalevolution. The proposed work introduces a geneticalgorithm based steganography method to protect againstthe RS attack in color images.II. LITERATURE SURVEYM.F.Tolba, M.A.Ghonemy and A.Taha  proposes analgorithm by which the information capacity can reach50% of the original cover image. It provides high qualityof stego image over the existing LSB based method.R. O., El.Sofy and H.H.Zayed  provide high hidingcapacity up to 48% of the cover image size. In this paper,they have tried to optimize these two main requirementsby proposing a novel technique for hiding data in digitalimages by combining the use of adaptive hiding capacityfunction that hides secret data in the integer waveletcoefficients of the cover image with the optimum pixeladjustment (OPA) algorithm.Ali Al-Ataby and Fawzi Al-Naima  propose a modifiedhigh capacity image steganography technique thatdepends on wavelet transform with acceptable levels ofimperceptibility and distortion in the cover image andhigh level of overall security.Souvik Bhattacharya, Avinash Prashad and GautamSanyal  incorporate the idea of secret key forauthentication at both the ends in order to achieve highlevel of security. In this paper, a specific image basedsteganography technique for communicating informationmore securely between two locations is proposed.H. S. Manjunatha Reddy and K. B. Raja  propose ahigh capacity and security steganography using discretewavelet transform (HCSSD). In this paper the two levelwavelet transform is applied as cover and payload. Thepayload wavelet coefficients are encrypted and fused withwavelet coefficients of cover image to generate stegocoefficients based on the embedding strength parametersalpha and beta.Elham Ghasemi, Jamshid and Brahram  propose anovel steganography scheme based on Integer WaveletTransform and Genetic Algorithm. Simulation resultsshow that the scheme outperforms adaptivesteganography technique based on integer wavelettransform in terms of peak signal to noise ratio andcapacity i.e. 35.17 dB and 50% respectively.T. C. Manjunatha and Usha Eswaran  use embeddingprocess stores up to 4 message bits in each integer co-efficient for all the transform sub-bands. This paperpresents a conceptual view of the digital steganography &exploits the use of a host data to hide a piece ofinformation that is hidden directly in media content, insuch a way that it is imperceptible to a human observer,but easily be detected by a computer.Amitav Nag, Sushanta Biswas, Debasree Sarkar andPartha Pratim Sarkar  present a technique for imagesteganography based on DWT. This paper presents anovel technique for Image steganography based on DWT,where DWT is used to transform original image (coverimage) from spatial domain to frequency domain. First,two dimensional Discrete Wavelet Transform (2-D DWT)is performed on a gray level cover image of size M × Nand Huffman encoding is performed on the secretmessages/image before embedding. Then each bit ofHuffman code of secret message/image is embedded inthe high frequency coefficients resulted from DiscreteWavelet Transform. Image quality is to be improved bypreserving the wavelet coefficients in the low frequencysub-band also.Yedla Dinesh and Addanki Purna Ramesh  perform amulti-resolution analysis and space frequencylocalization. As compared to the current transformdomain data hiding methods this scheme can provide anefficient capacity for data hiding without sacrificing theoriginal image quality.Saddaf Rubab and M.Younus  derive a new algorithmto hide our text in any colored image of any size usingwavelet transform. It improves the image quality andimperceptibility. Their method sustains the securityattacks. This new method gives better invisibility andsecurity of communication. This method provides doublesecurity by involving blowfish, which satisfies the need ofimperceptibility.S.Priya and A.Amsaveni  give LSB based edgeadaptive image steganography. Edge adaptivestenography on frequency domain improves security and
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________396IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________image quality compared to the edge adaptive stenographyon spatial domain.Rastislav Hovancak, Peter Foris and Dusan Levicky propose a new method of steganography technique basedon DWT transform. The proposed method has ability tohide secret message in a digital image. The secretmessage is embedded into the image by changing waveletco-efficient. The quality of the stego image of theproposed method is very close to that of the original one.Arezoo Yadollahpour and Hossein Miar Naimi proposed a steganalysis technique using auto-correlationcoefficients in colour and grayscale images. They suggestthat insertion of secret message weakens the correlationbetween the neighbour pixels and thereby enabling one todetect the message.Fridrich et al.  proposed an effective steganalysistechnique popularly known as RS steganalysis, which isreliable even in the detection of non-sequential LSBembedding in digital images.Andrew D Ker  has proposed a general framework forstructural steganalysis of LSB replacement for detectionand length estimation of the hidden message. He hassuggested the use of previously known structuraldetectors and recommended a powerful detectionalgorithm for the aforementioned purpose.Tao Zhang and Xijian Ping  have proposed asteganalysis method for detection of LSB steganographyin natural images based on different histograms. Thismethod ensures reliable detection of steganography andestimate the inserted message rate. However, this methodis not effective for low insertion rates.Fridrich and Goljan  have considered manysteganalysis techniques and proposed a steganalysistechnique based on image‟s biplanes correlation. Theystate that LSB plane can be estimated from 7 planes out of8 planes in a pixel of the image. They feel that theperformance of the suggested steganalysis methodreduces as the LSB plane‟s content is further randomized.Kong et al.  proposed a new Steganalysis approachbased on both complexity estimate and statistical filter. Itis based on the fact that the bits in the LSB plane arerandomized when secret bits are hidden in LSB plane.Amirtharajan et al.  proposed a novel and adaptivemethod for hiding the secret data in the cover image withhigh security and increased embedding capacity. Theyfeel that by using this method the receiver does notrequire the original image to extract the information.Umamaheswari et al.  proposed analysis of differentsteganographic algorithms for secure data hiding. Theyrecommend compressing the secret message andencrypting it with receiver public key along with the stegokey. They have analyzed different embedding algorithmsand used cryptographic technique to increase the security.Taras Holotyak e.t. al  propose a new method forestimation of the number of embedding changes for non-adaptive ±k embedding in images. The same author has also advocated a new approach to blind steganalysis,based on classifying higher-order statistical featuresderived from an estimation of the stego signal in thewavelet domain.Agaian and Perez  propose a new steganographicapproach for palette-based images. This recently approachhas the advantage of secure data embedding, within theindex and the palette or both, using special scheme ofsorting. The presented technique also incorporates the usecolor model and cover image measures, in order to selectthe best of the candidates for the insertion of the stegoinformation.Chen and Lin  propose a new steganographytechnique which embeds the secret messages in frequencydomain to show that the PSNR is still a satisfactory valueeven when the highest capacity case is applied. Bylooking at the results of simulation, the PSNR is still arelaxed value even when the highest capacity is applied.This is due to the different characteristics of DWTcoefficients in different sub-bands. Since, the mostessential portion (the low frequency part) is keptunchanged while the secret messages are embedded in thehigh frequency sub-bands (corresponding to the edgesportion of the image), good PSNR is not a imaginaryresult. In addition, corresponding security is maintainedas well since no message can be extracted without the“Key matrix” and decoding rules.Kathryn Hempstalk  investigates using the cover‟soriginal information to avoid making marks on the stego-object, by hiding the basic files of electronic reside digitalcolor images. This paper has introduced two imagesteganography techniques, FilterFirst and BattleSteg.These two techniques attempt to improve on theeffectiveness of hiding by using edge detection filters toproduce better steganography.Wang and Moulin  provided that the independent andidentical distributed unit exponential distribution model isnot a sufficiently accurate description of the statistics ofthe normalized periodogram of the full-frame 2-D imageDFT coefficients.Park e.t. al  proposed a new image steganographymethod to verify whether the secret information had beenremoved, forged or altered by attackers. This proposedmethod covers secret data into spatial domain of digitalimage. In this paper, the integrity is verified fromextracted secret information using the AC coefficients ofthe discrete cosine transform (DCT).Ramani, Prasad, and Varadarajan  proposed an imagesteganography system, in which the data hiding(embedding) is realized in bit planes of subband waveletscoefficients obtained by using the Integer WaveletTransform (IWT) and Bit-Plane ComplexitySegmentation Steganography (BPCS).Farhan and Abdul  have presented their work inmessage concealment techniques using image basedsteganography.Anindya e.t. al  presented further extensions of yetanother steganographic scheme (YASS) which is amethod based on embedding data in randomized locationsso as to resist blind steganalysis. YASS is a technique of
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________397IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________JPEG steganographic that hides data in the discrete cosinetransform (DCT) coefficients of randomly chosen imageblocks.Adnan Gutub e.t. al.  depicts the random pixelmanipulation methods and the stego-key ones in thepropose work, which takes the least two significant bits ofone of the channels to indicate existence of data in theother two channels. This work showed good resultsespecially in the capacity of the data-bits to be hiddenwith relation to the RGB image pixels.Mohammed and Aman  used the Least SignificantBits (LSB) insertion method to hide data within encryptedimage data.Aasma Ghani Memon e.t. al.  provides a new horizonfor safe communication through XML steganography onInternet.Zaidan e.t. a.l.  has presented a model for protectionof executable files by securing cover-file withoutlimitation of hidden data size using computation betweencryptography and steganography.Vinay Kumar and Muttoo  have discussed that graphtheoretic approach to steganography in an image as coverobject helps in retaining all bits that participate in thecolor palette of image.Wang e.t. al.  presented a new steganography basedon genetic algorithm and LSB.In recent research works few algorithms have beenproposed which consist of the marginal statistics that arepreserved for achieving more security. Previous methodshave less data hiding capacity and security againstReversible Statistical attack. As we increase the secretdata length distortion increases in the final stego image ascompared with cover image. All the previous worksprovide the basic idea to hide the data behind the imageby using LSB substitution. There is no idea discussedabout the increasing capacity of data so no effect onimage and how to ban the RS attack. This is a criticalissue in steganography model that how we increase thehiding capacity of an image or cover media without anydistortion in the image quality and how to protect themethod against the RS attack.III. PROPOSED SYSTEM ARCHITECTUREDesign is a necessary phases of code development. Thedesign is a methodology throughout that a systemorganization is established that is able to satisfy thesensible and non-functional system wants. Large Systemsare divided into sub-systems that offer few connected setof services. The design process output is an architecturedescription. With regular analysis and improvement instyle of algorithmic program, steganography is taken as asignificant meaning to cover information and additionallythe current work appears that it is efficient in hiding alarge amount of information. GA is applied to realizeassociate optimum mapping function to cut back the errordistinction between the input cover and the stego imageand use the block mapping methodology to preservenative image properties and to cut back the complexnessof algorithmic program. Optimal pixel adjustment processis applied to increase the hiding capability of thisalgorithmic program compared to other existing systems.In this high level system design the whole system designand development is to be administered. The systemdevelopment with the correct sequence and therefore thesynchronization with the all connecting modules measureaiming to be lined within the tactic of high level comingup with. The Genetic algorithm implementation is inaddition one of the necessary steps for the high levelsystem design. During this development method the GAhas been used for the RS analysis.Design issuesThe proposed work presents a replacementsteganographic technique in order to embed large amountof data in colored images whereas keeping the activitydegradation to a minimum level using integer wavelettransform (IWT) and Genetic algorithm (GA). Thistechnique permits concealment of a data in uncompressedcolor image. Our motivation to cover data in images is toprovide security to images that contain crucial data.Proposed approach relies on LSB technique which is ableto replace more than one bit from every pixel to coversecret message, but the security of the secret data can beimproved by combining the least significant bit andwavelet transform. The aim of the design is to plan thesolution of a given problem by the document needs. It isthe beginning in moving from drawback to the solutiondomain. The design of the system is the most vital issueaffecting the quality of the computer code package andcontains a major impact on the coming phases such astesting and maintenance. The proposed work is basicallyexperimental test-bed for analysis of RS-attack using LSBfurthermore as genetic algorithm. So the design to bethought of during this work ought to be a frameworkapplication in MATLAB in integrated developmentsetting considering all the parameters to protect the datausing advance steganography.Assumptions and dependenciesThe primary assumption of the work is that the user istaking the input of original image and not from anyprocessed or manipulated image.The user is predicted to use the standard cryptographyalgorithmic program in an exceedingly most securesystem and network.The basic dependency of the work is to run theapplication, user needs the MATLAB setting and touse application and appraise its basic conception, userneeds associate noise free image and knowledge inplain text format solely.ConstraintsThe application relies on optimization using genetic rulewithin the current steganographic applications. Herelimitation is that its been found that whenever a pictureinput is subjected to such forms of process then there isloss of actual quality of image. So on resist RS analysis,the impact on the relation of pixels must be stipendiarywhich cannot be achieved by adjusting totally different bitplanes. The implementation procedure may be
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________398IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________unworkable in non theoretical application. Therefore toovercome this limitation, GA is applied to calculate thehigher adjusting mode that the image quality is notdegraded.Proposed system architectureThe planned work ensures the safety against the RSanalysis. The application should be designed in such away so as to overcome all the limitation considered withinthe previous analysis work. The present aim is to style thearchitecture of the planned work which dependscompletely on a sturdy process of safeguarding the inputto the application. This strategy incorporatesimplementing least necessary bit for embedding the keymessage of the quilt image. Successive issue could be theloss of quality of the image and therefore the planning isdone for safeguarding the standard of the image which isachieved by implementing Genetic algorithmic rule. It is away of search employed in computing to search out exactor approximate solutions to optimization and searchissues.This work presents a completely unique steganographytechnique which will ultimately increase the capability ofdata embedding and therefore the imperceptibility of theimage after embedding. The proposed system architectureis highlighted as below:Fig. 4.1. Proposed system architectureFig. 1. Proposed System ArchitectureThe complete process can be expressed as follows:Fig. 2. Complete flow of proposed workThe above mentioned figure represents the general systemfunctionalities and the real operative steps of thedeveloped design. In the processing, the program helps soas give a program to handle the developed model and toaccess the developed module. At the origination, thecover image is selected for embedding the message. Thenthe text data is to be selected so, as to accomplish themotive of steganography the stego key applied so at theopposite terminal the message can be retrieved by thesame key. Once the Key is provided, the real applicationdevelopment for the RS analysis will be started with thestrong GA improvement. In this technique, the message isto be embedded in cover image. Genetic algorithm isplaying an important role for embedding more and moredata in the image. In the architecture of the developedsystem the integer to integer wavelet transform is applied.Once the message is embedded into the image file, thenembedding the image is again recovered so that it is nowable to be transmitted over the channel. On the otherhand, at the receiver terminal or at the extraction terminalwith the accurate stego key, the message is retrievedaccurately.IV. PROPOSED WORKDetailed design of the proposed steganography givesexhaustive image of the foremost parts described in thesystem design. Meantime this chapter describes the detaildesign of the system. In this section details and flow chartof each module has been described. The structure chartshow control flow, the useful descriptions of that areconferred in the flow chart diagrams.Module specificationSelection MutationInput DataInput Cover ImageSecret Text MessageInteger Wavelet Transform (IWT)BlockingGenetic Algorithm (GA)ChromosomeInitializationCrossoverPerform 2D-IWTEvaluate Regular and Singular Block Values(RM, R-M, SM, S-M)Block FlippingPerform RS AnalysisGraphical User InterfaceSelect the Input Cover ImageSelect the Secret Text to be EmbeddedInsert the Secret KeyGA Design Based RS ParametersMessage EmbeddingInverse Wavelet TransformFitness FunctionEmbedded MessageOPAP Algorithm2D Inverse IWTMessage ExtractionRS AnalysisMapping Function
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________399IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________The proposed model is prepared by using twofundamental modules:A) Embedding module: The main task of this module is toembed a secret text within the cover colored image usingencryption key. The complete cover image is divided into8x8 blocks before any further processing. The frequencydomain representation of the respective created blocks isestimated by two dimensional Integer wavelet transform inorder to accomplish 4 sub bands LL1, HL1, LH1, and HH1.This way 1 to 64 genes are generated containing the pixelsnumbers of each 8x8 blocks of the mapping function. Themessage bits in 4-LSBs coefficients of IWT in each pixelaccording to mapping function are embedded. Fitnessevaluation based, Optimal Pixel Adjustment Process on theImage is applied. At last, inverse 2D IWT is computed inthis module in order to generate the stego image.B) Extraction module: The main task of this module is theextraction of the actual secret text from the stego image tounderstand the effectiveness of process of messageembedding. It takes the stego image as input with key fordecrypting the hidden message from the stego image. Oncethe data has been transmitted over the communicationchannel and when the receiver receives the embedded imagefile, then it becomes necessary to again segment the imagedata and then take out the text data available at the spacecovered by the text data at the time of message embedding.The extraction can be summarized in a simple sentence as totake out the data that has been embedded.Genetic algorithm utilization processA Structure Chart (SC) in software engineering andorganizational theory is a chart, which shows thedeviation of the system configuration to the lowestmanageable levels. Steganalysis is the art and science ofdetecting messages hidden using steganography; this isanalogous to cryptanalysis applied to cryptography. Theobjective of steganalysis is to find suspected packages,identified that they have a payload encoded into them ornot, and, if it is possible, then resolve that payload.Unlike cryptanalysis, where it is obvious that intercepteddata contains a message (though that message isencrypted), generally steganalysis begins with a pile ofsuspect data files, but few information about which of thefiles, if anyone, contain a payload of information. Thesteganalyst is usually something of a forensic statistician,and should begin by minimizing this set of data files(which is often quite large; in a lot of cases, it may be thewhole set of files on a computer) to the subset most likelyto have been altered. In computing, the smallest amountof important bit (LSB) is that the bit position in a verybinary number giving the units price, that is, decisivewhether or not the quantity is even or odd. The LSB isusually remarked because the right-most bit, as a result ofthe convention in number system of writing lesser digitany to the correct. It is analogous to the smallest amountfigure of a decimal number, that is that the digit within theones (right-most) position. A genetic algorithm (GA) is asearch technique used in computing to find exact orapproximate solutions to optimization and searchproblems. Genetic algorithms are divided as world searchheuristics. Genetic algorithms are a basic category ofevolutionary algorithms (EA) that use techniquesgalvanized by organic process biology like inheritance,mutation, selection, and crossover.The following figure represents the structural chartrepresentation for the proposed system development. Hereit represents the overall processing and the step by steppresentation of the proposed work.Fig. 3. GA utilization processModule designThis section contains a detailed description ofcomponents of software, components of low-level andother sub-components of the proposed work. ModuleNoNoNoYesSecretTextInputImageFinalImageEmbeddedGABlockingLabelsChromosomeSelectionCheckLabelsReproductionMutationCrossoverCrossover>2RSConditionNextBlock
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________400IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________design helps for the implementation of the modules.Module‟s input requirements and outputs generated by themodules are described in this phase.Data embeddingThis is the process flow diagram for data embeddingmodule to illustrate the initiation of security featuresalong with implementation of IWT and GeneticAlgorithm. The main purpose of this application is toshow the flow of data embedding operation involved inthe process. The frequency domain representation of therespective created blocks is estimated by two dimensionalInteger wavelet transform in order to accomplish 4 subbands LL1, HL1, LH1, and HH1. 1 to 64 genes aregenerated containing the pixels numbers of each 8x8blocks as the mapping function. The bits of message in 4-LSBs IWT coefficients each pixel according to mappingfunctions are embedded. According to fitness evaluation,Optimal Pixel Adjustment Process applied on the Image.At the end, inverse 2D IWT is computed in this module inorder to generate the stego image. The input for thisprocessing is basically a cover image and user textmessage for embedding purpose. Stego image isgenerated as a output after this process. This moduleinteracts with all the components of the applicationresponsible for selection of parameters for performingencryption.Fig. 4. Flow chart of the data embedding processData extractionFigure 5 shows the process flow diagram for messageextraction module to illustrate the decryption hidden textin the stego image. The main purpose of this applicationis to show the flow of message extraction operationsinvolved in the process. This algorithm basically takes theinput of the generated stego image from the embeddingprocess and applies IWT along with decryption key toextract the secret text which has been hidden inside thestego image. The input for this processing is basically astego image and decryption key for message extractionpurpose. Original user text is generated as output afterthis process. This module mainly interacts with thepreviously implemented message embedding process forperforming extraction.Fig. 5. Flow chart of the data extraction processLSB implementationFigure 6 shows the flow chart will show the section whereLSB is implemented. The major operation takes placewhen the application starts getting the size of the coverimage and then it creates a tree structure for ease incomputation. After it gets filter value of the pixels, wherethe application start the filter and configure the startingand ending bits, that last set the match image. Afterperforming this operation, LSB algorithm will beimplemented in the cover image, where the pixels valuesof the stego-image are modified by the genetic algorithmto keep their statistic characters. Inputs are embeddingoriginal message with cover image. Output of the processis actual implementation of LSB algorithm. This moduleinteracts with LSB module and genetic algorithm alongwith input files of cover image.StartDivide Image in 8x8BlocksStego ImageExtract CoefficientLSB ImplementationPixel SequenceSecret KeyActual DataStopStartTake InputCoverImageTake SecretText DataIWT ProcessDivide theInput Imagein 8x8BlocksGather allCoefficientsStore Coff.4 Sub BandsPermutations PixelInformation(EachBlock)Mapping Func.LSB ProcessFitness Func.OPAP2D-I-IWTStego ImageStop
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________401IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________Fig. 6. Flow chart of the LSB implementationWavelet applicationsIn mathematics, a wavelet series is an illustration of asquare-integrable real number or complex number orcomplex valued function by a certain orthonormal seriesgenerated by a wavelet.Wavelet transformWavelet domain techniques are becoming very popularbecause of the developments in the wavelet stream in therecent past years. Wavelet transform is employed toconvert a spatial domain into frequency domain. Theemployment of wavelet in image stenographic model liesin the fact that the wavelet transform clearly separates thehigh frequency and low frequency information on a pixelby pixel basis. A continuous wavelet transform (CWT) isused to divide a continuous-time function into wavelets.Integer wavelet transformThe proposed algorithm employs the wavelet transformcoefficients to embed messages into four subbands of twodimensional wavelet transform. To avoid problems withfloating point precision of the wavelet filters, we usedInteger Wavelet Transform. The LL subband in the caseof IWT appears to be a close copy with smaller scale ofthe original image while in the case of DWT the resultingLL subband is distorted (figure 7) . Thus IntegerWavelet Transform (IWT) is preferred over DiscreteWavelet Transform (DWT).(a) Lena image and analyze inwavelet domain(b) One level 2D-DWT in subband LL (c) One level 2D-IWT insubband LLFig. 7. Comparison of LL subband for 2D-DWT and 2D-IWTIn 2D IWT transform, first apply one step of the onedimensional transform to all rows and then repeat towhole columns. This decomposition outputs into fourclasses or band coefficients. The Haar Wavelet Transformis the easiest of all wavelet transform. In this transform,the low frequency wavelet coefficient are generated byaveraging the two pixel values and high frequencycoefficients are generated by taking half of the differenceof the same two pixels. The 4 bands produced are (i)Approximate band (LL), (ii) Vertical Band (LH), (iii)Horizontal band (HL), (iv) Diagonal detail band (HH).The approximation band consists of low frequencywavelet coefficients, which have important parts of thespatial domain image. The last band consists of highfrequency coefficients, which contain the edge details ofthe spatial domain image. This IWT decomposition of thesignal continues until the desired scale is achieved .Two-dimensional signals, like images, are converted using the2D IWT. The two-dimensional IWT operates in the samemanner, with only minor variations from the one-dimensional transform. Given a two-dimensional array ofsamples, the rows of the array are processed first withonly one level of decomposition. This essentially dividesthe array into two vertical halves; with the first half takingthe average coefficients, while the second vertical halfstores the detailed coefficients. This process is againperformed with the columns, resulting in 4 sub bandswithin the array defined by filter output.Integer wavelet transform through lifting schemeThe lifting scheme is for both designing wavelets andperforming the discrete wavelet transforms. Basically it isworthwhile to merge these steps and design the waveletfilters while performing the wavelet transform. Themethod was introduced by Wim Sweldens . Thelifting scheme is an algorithm to calculate wavelettransforms in an effective way. It is also a generictechnique to create so-called second-generation wavelets.They are much more flexible and can be used to defineLL LHHL HHStartCover ImageSizePixelCapacityHalfwayComputationFilter valueMatch ImageLSBImplementationSet FilterStart Bits &End BitsLSB=Match StopDWT
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________402IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________wavelet basis on an interval or on an irregular grid, oreven on a sphere. The wavelet lifting scheme is a methodfor decomposing wavelet transform into a set of stages.An advantage of lifting scheme is that they do not requiretemporary storage in the calculation steps and require lessno of computation steps. The lifting procedure consists ofthree phases: (i) split phase, (ii) predict phase and (iii)update phase.Fig. 8. Lifting scheme forward wavelet transformationSplitting: Divide the signal x into even samples and oddsamples:xeven : si ← x2i, xodd : di ← x2i+1Prediction: Analyze the odd samples using linearinterpolation:di ← di − (si+si+1)/2Update: Update the even samples to maintain the meanvalue of the samples:si ← si + (di−1+di)/4The output from the s channel provides a low pass filteredversion of the input where as the output from the dchannel provides the high pass filtered version of theinput. The inverse transform is obtained by reversing theorder and the sign of the operations performed in theforward transform .Fig. 9. Lifting scheme inverse wavelet transformationLifting scheme Haar transformIn the lifting scheme version of the Haar transform,predicts that the odd element will be equal to the evenelement. The difference among the predicted value (theeven element) and the actual value of the odd elementreplaces the odd element. For the forward transformiteration j and element i, the new odd element, j+1,iwould be: oddj+1,i = oddj,i − evenj,i. In the lifting schemeversion of the Haar transform the update step replaces aneven element with the average of the even /odd pair (e.g.the even element si and its odd successor si+1) is evenj+1,i =(evenj,i+oddj,i)/2 . The original value of the oddj,i elementhas been replaced by the difference between this elementand its even predecessor. The original value is :oddj,i =evenj,i + oddj+1,i.Substituting this into the average, we getevenj+1,i = (evenj,i+evenj,i+oddj+1,i)/2 .Genetic algorithm based steganography methodThe proposed method embeds the message inside thecover image with the minimal distortion. Use a mappingfunction to LSBs of the cover image according to thecontent of the message. Genetic Algorithm is used to finda mapping function for all the image blocks. Block basedstrategy preserve local image property and reduces thealgorithm complexity as compared to single pixelsubstitution. The genetic algorithm optimizes the imagequality and security of the data.Chromosome designIn our GA method, a chromosome is encoded as an arrayof 64 genes containing permutations 1 to 64 that point topixel numbers in each block. Each chromosome producesa mapping function (figure 10).59 47 1 33 …………….. 41 16 9 60Fig. 10. Chromosome with 64 genesEach pixel in a block is considered as a chromosome.Some chromosomes are considered for forming an initialpopulation of the first generation in genetic algorithm.Several generations of chromosomes are created to selectthe best chromosomes by applying the fitness function toreplace the original chromosomes. Reproductionrandomly duplicates some chromosomes by flipping thesecond or third lowest bit in the chromosomes. Severalsecond generation chromosomes are generated. Crossoveris applied by randomly selecting two chromosomes andcombining them to generate new chromosomes. This isdone to eliminate more duplication in the generations.Mutation changes the bit values in which the data bit isnot hidden and exchanges any two genes to generate newchromosome. Once the process of selection, reproductionand mutation is complete, the next block is evaluated.GA operationsMating and mutation functions are applied on eachchromosome. The mutation process causes the inversionof some bits and produces some new chromosomes, then,we select elitism which means the best chromosome willsurvive and be passed to the next generation.Fitness functionSelecting the fitness function is one of the most importantsteps in designing a Genetic Algorithm based method.Whereas Genetic Algorithm aims to improve the imagequality, Peak Signal to Noise Ratio (PSNR) can be anappropriate evaluation test.The fitness function enables to optimize the value throughseveral iterations. Fitness is calculated by the probabilityof regular and singular groups when positive flipping andnegative flipping is applied. Ultimately, the stego-imageundergoes RS analysis and the values between originaland stego-image are compared.Block flippingOdd ValuesEven ValuesSplit Predict Update+-OddValuesEvenValuesMergeUpdate Predict+-
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________403IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________RS steganalysis classifies block flipping into three types.They are positive flipping F1, negative flipping F-1, andzero flipping F0. RS steganalysis analyses three primarycolors namely red, green and blue individually for colorimages. Initially, the image is divided into several blocks.Subsequently, flipping functions such as positive flippingand negative flipping are applied on each block of pixels.Later, the variations between original and flipped blocksare calculated. Based on the variation results, the blocksare categorized into regular and singular groups. Let RMdenote relative number of regular group and SM denoterelative numbers of singular groups. According to thestatistical hypothesis of the RS steganalysis method in atypical image, the expected value of RM is equal to that ofR−M, and the same is true for SM and S−M:RM R−M and SM S−MWith application of positive flipping, RM denotes regulargroup and SM is singular group. Similarly, R-M and S-M areregular and singular group when negative flipping isapplied. The difference between regular groups, RM andR-M and the difference between singular groups, SM and S-M increases with the increase in length of the secretmessage.V. IMPLEMENTATION AND EXPERIMENTALRESULTS DISCUSSIONThe important phase of a research work is itsimplementation which shows the actual direction ofimplementing the scenario, methods and step by stepdevelopment. The implementation part of anydevelopment is the implementation part as the same yieldsthe ultimate solution, which solves the matter in hand.The phase of implementation involves the actualmaterialization of the ideas, which are show in thedocument analysis and are developed in the phase ofdesign. Implementation should be the best mapping of thedesign document in a suitable programming language inorder to achieve the necessary final product. Usually theproduct is ruined due to incorrect programming languageadopted for implementation or unsuitable method ofprogramming. It is better for the phase of coding to bedirectly connected to the design phase in the sense if thedesign is in terms of object oriented terms thenimplementation should be preferably carried out in aobject oriented way. The implementation of the systemdeveloped has been performed on the MATLAB softwareplatform.ImplementationImplementation of proposed steganography application isalways preceded by important decisions regardingselection of the platform, the language used, etc. Thesetypes of decisions are often influenced by several factorssuch as real environment in which the system works, thespeed required, the security issues, and implementationrelated details. These major implementation decisions arethere that have been made before the implementation ofthe work.Proposed work implementation requirementsThe implementation of the proposed work requires aninput cover image with a data file for performing themessage embedding process. However the softwarerequirements for performing the implementation are:MATLAB 22.214.171.1249 (R2010a)Microsoft windows XP.NET framework 3.5Guidelines to perform codingThe following guidelines have been used during theimplementation of the proposed work:Initialize local variables and all pointers initialized tothe defined values or NULL.Use tracing statements at critical points in the code.For all the data types, type definitions are used.All the message formats are stored in header file.All the functions should not exceed more than 100lines.Function pointers are not used.All the codes should be properly indented.Use conditional compilation statements, whereverrequired.Implementation of algorithmData embedding algorithmThe proposed method for data hiding comprises of thefollowing:Take the input standard cover image.Take the secret text message.Apply the secret key (in digits only).Perform the Integer Wavelet Transform of the inputcover image using lifting scheme.Add primal ELS to the lifting scheme.Perform integer lifting wavelet transform on image.Divide the input cover image in 8x8 blocks.Select any of the wavelet coefficients (redundantcoefficients) from the obtained high frequencycoefficients.Generate 64 genes containing the pixels numbers ofeach 8x8 blocks as mapping function.Initialize empty matrix to store the wavelet values.Obtain 8x8 blocks for R G B.Concatenate all coefficients together.Store the coefficient in new image.Embed in K-LSBs IWT coefficients in each pixelaccording to mapping function.Select any one of the pixels from RGB.Now the selected coefficients are processed to make itfit for modification or insertion.Fitness evaluation is performed to select the bestmapping function.The secret message plus the message length isembedded into the processed coefficients.This modified coefficient is now merged with theunmodified coefficients.Calculate embedded capacity.Apply Optimal Pixel Adjustment Process on theimage.
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________404IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________Convert image to binary.Finally, the inverse 2D IWT on each 8x8 block isapplied to obtain the Stego image.Stego image to be obtained.Data extraction algorithmThe proposed method for data extraction comprises of thefollowing:Take the desired stego image.Apply the same secret key as given in embeddingprocess.Divide the stego image into 8x8 blocks.Extract the transform domain coefficient by 2D IWTof each 8x8 blocks.Find the pixel sequences.Select the desired pixels for process.Extract K-LSBs in each pixel.Process the selected pixels coefficient to make it fit,for extraction.Now extract the message length and the secretmessage from these processed coefficients.Secret message to be obtained.RS-analysis algorithmThe proposed method for RS analysis comprises of thefollowing:Create function for non-positive flipping (Fn).Create function for non-negative flipping (Fp).Change LSB as per flipping.Initialize Relative number of regular block afterpositive flipping (R+) = 0.Initialize Relative number of Singular block afterpositive flipping (S+) = 0.Divide Stego Image into 8x8 blocks.For a modified block B, apply the non-positiveflipping F− and the non-negative flipping F+ on theblock. The flipping mask M+ and M− are generatedrandomly. The result is B+ and B−.Estimate F (B+), F (B−) and F (B).Define four variables to divide the blocks bycomparison of F (B+), F (B−) and F (B).Initially P+R = 0, P+S = 0, P-R = 0 and P-S = 0.Do the following steps for 100 timesFor nn = 1:100Apply the non-positive flipping F-.Fn = non_positive_flipping (B).Apply non-negative flipping F+.Fp = non_negative_flipping (B).Calculate f (B0+), f (B0-) and f (B).C = calculate_correlation (B).Correlation for non positive flipping.Cn = calculate_correlation (Fn).Correlation for non positive flipping.Cp = calculate_correlation (Fp).Estimate P+R, the count of the occurrence when theblock is regular under the non-negative flipping.Estimate P+S, the count of the occurrence when theblock is singular under the nonnegative flipping.Estimate P−R, the count of the occurrence when theblock is regular under the non-positive flipping.Estimate P−S, the count of the occurrence when theblock is singular under the non-positive flipping.If Cn>C, then increase P−R (Regular).P−R = P−R +1.Else, increase P−S (Singular).P−S = P−S +1.EndIf Cp>C, then increase P+R (Regular).P+R = P+R +1.Else, increase P+S (Singular).P+S = P+S +1.EndCompare P+R to P+S and P−R to P−S, the block‟s labelare determined, str = .If P+R / P+S >1.8, then str = R+.disp (R+), Label of the block „R+‟.Rp = Rp+1.EndIf P+S / P+R > 1.8, then str = S+.disp (S+), Label of the block S+.Sp = Sp+1.EndIf P−R/P−S > 1.8, then str = [str R-].disp (R-), Label of the block R-.Rm = Rm+1.EndIf P−S / P−R > 1.8, then str = [str S-].disp(S-), Label of the block S-.Sm = Sm+1.EndAt last, the blocks are categorized into 4 groups(R+R−), (R+S−), (S +R−), (S +S−).Reject the block which doesn‟t fall in 4 groups.Now use genetic algorithm for minimizing R- block.The blocks, which are not included in the 4 categories, arenot processed in following steps. Compared to theoriginal image, the values of R+ R− and S+ S− blocks areincreased in the stego-images. This phenomenon can bedetected by the RS analysis. The main aim of theproposed algorithm is to decrease the amount of R−blocks. Therefore genetic algorithm is deployed to adjustthem to maintain the visual quality of image as given infollow section.Optimization technique or genetic algorithmThe proposed method for genetic algorithm comprises ofthe following:Perform Chromosome Initialization Steps.From the first pixel, select every 4 pixels.B1 = B (:)crossover = 0.Initialize Alpha as 0.88.
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________405IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________For kk = 1: length (B1) - 2.do Chrom = B1 (kk:kk+2).Initialize maximum Fitness as 0.Flip second lowest bit randomly for number of time.For kk1 = 1:100Cp = non_negative_flipping (Chrom).Cn = non_positive_flipping (Chrom).C = calculate_correlation (Chrom).Cn = calculate_correlation (Cn).Initialization, e1 = 0 and e2 = 0.If Cn<C, then e1 = 1.EndIf Cp > C, then e2 = 1.EndPSNR = snr (Chrom-Cn).fitness = alpha*(e1+e2)+PSNR.If fitness>maxfitness, then maxfitness = fitness.Chrommax = Cp.crossover = crossover+1.EndReplace chromosome with new one.B1 (kk:kk+2) = Chrommax.Calculate P-s and P-r.For qq = 1:100Apply the non-positive flipping F-.Fn = non_positive_flipping(B1).Calculate f (B0+), f (B0-) and f (B).C = calculate_correlation (B1).Correlation for non positive flipping.Cn = calculate_correlation (Fn).If Cn>C, then Regular.P-R = P-R +1.Else, Singular.P-S = P-S +1.EndIf P-S > P-R, then disp (block is successfully adjusted).EndIf crossover>2, then breakEndP+R = 0, P+S = 0, P-R = 0, P-S = 0.Do the following steps for 100 timesFor nn = 1:5Apply the non-positive flipping F-.Fn = non_positive_flipping (B).The non-negative flipping F+.Fp = non_negative_flipping (B).Calculate f (B0+), f (B0-) and f (B).C = calculate_correlation (B).Correlation for non positive flipping.Cn = calculate_correlation (Fn).Correlation for non positive flipping.Cp = calculate_correlation (Fp).If Cn>C, then Regular.P-R = P-R +1.Else, Singular.P-S = P-S +1.EndIf Cp>C, then Regular.P+R = P+R +1.Else, Singular.P+S = P+S +1.Enddiff1 = abs (P+R – P-R).diff2 = abs (P+S – P-S).If difference is more than 5% then.If diff1>0.05*diff2.Successful then replace.I (ii:ii+7,jj:jj+7) = reshape (B1,8,8).Break the loop and go for next block.In the proposed technique, the blocks are labeled beforethe adjustment. Thus, the computational complexity isminimized. Genetic method use avoids the exhaustivesearching and the algorithm is easy to be implemented.Proposed work implementationThe proposed implementation of RS-analysis usinggenetic algorithm for the robust security in Steganographyapplication is done on standard 32-bit windows OS with1.84 GHz processor and 2 GB RAM. The method isapplied on 512x512 colored images “Lena” and “Baboon”as shown in Figure 11.a) Lena (JPG, 512x512) b) Baboon (JPG, 512x512)Fig. 11. Input cover imagesExperimental result analysis and discussionThe proposed work is done on 2 set of data image asshown in previous section. Both cover images haveutilization of 100% and their respective accomplishedresults of reversible statistical analysis are as follows:TABLE 1VARIOUS VALUES FOR LENA IMAGEFor Lena Initial ValueAfterEmbeddingAfterOPAPRm-R-m 0.0097783 0.0076353 0.0057934Sm-S-m 0.0029662 0.011807 0.0093702TABLE 2VARIOUS VALUES FOR BABOON IMAGEForBaboonInitialValueAfterEmbeddingAfterOPAPRm-R-m 0.0059805 0.0076353 0.0056089Sm-S-m 0.0076634 0.011807 0.0023989The tables 1 and 2 have shown the values of |Rm-R-m| and|Sm-S-m| that represent the RS-steganalysis on the regularand singular block. It can be seen that the value of |Rm-R-m| and |Sm-S-m| increases from initial value beforeembedding and after embedding that exhibits a strong
International Journal on Recent and Innovation Trends in Computing and CommunicationVolume: 1 Issue: 4 394 – 408___________________________________________________________________________408IJRITCC | MAR 2013, Available @ http://www.ijritcc.org___________________________________________________________________________University of Sharjah, Sharjah, U.A.E. 18 – 20 March2008. Mohammad Ali Bani Younes and Aman Jantan, “A NewSteganography Approach for Image Encryption Exchangeby Using the Least Significant Bit Insertion”, IJCSNSInternational Journal of Computer Science and NetworkSecurity, VOL.8 No.6, June 2008. Aasma Ghani Memon, Sumbul Khawaja and AsadullahShah, “STEGANOGRAPHY: A new horizon for safecommunication through XML”, Journal of Theoreticaland Applied Information Technology, 2008. A.A.Zaidan, Fazidah.Othman, B.B.Zaidan , R.Z.Raji,Ahmed.K.Hasan and A.W.Naji, “Securing Cover-FileWithout Limitation of Hidden Data Size UsingComputation Between Cryptography and Steganography”,Proceedings of the World Congress on Engineering 2009Vol I WCE 2009, July 1 - 3, 2009, London, U.K. Vinay Kumar, S. K. Muttoo, “Principle of Graph TheoreticApproach to Digital Steganography”, Proceedings of the3rd National Conference; INDIACom-2009. Shen Wang, Bian Yang and Xiamu Niu, “A SecureSteganography Method based on Genetic Algorithm”,Journal of Information Hiding and Multimedia SignalProcessing, Volume 1, Number 1, January 2010. W. Sweldens, The lifting scheme: A construction of secondgeneration wavelets, SIAM J. Math. Anal., 29:511–546,1997.