Documents

A Secure Spatial Domain Image Steganography Using Genetic Algorithm and Linear Congruential Generator

Description
free for all
Categories
Published
of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  A Secure Spatial Domain ImageSteganography Using Genetic Algorithmand Linear Congruential Generator Pratik D. Shah and R.S. BichkarAbstract   Signi 󿬁 cant increase in data traf  󿬁 c over the Internet has given rise tomany data security issues. Steganography is a technique which is used to hide theexistence of secret communication. Hence, it is extensively used to solve the issuesrelated to data security. In this paper, a secure and lossless spatial domain imagesteganography technique is proposed. Stream of secret data is hidden in quarter part of image by identifying suitable locations to hide 2 bits of secret data in each pixel,resulting in generation of coef  󿬁 cients corresponding to the location of match. Thesecoef  󿬁 cients are hidden in remaining part of image using LSB replacement steganography. Genetic algorithm is used to  󿬁 nd best possible location to hide thesecoef  󿬁 cients in the image, making the proposed technique very secure and almost impossible to extract secret data from it. The result of the proposed technique iscompared with LSB replacement steganography where in same amount of secret data is embedded. It is observed that the proposed technique is much superior ascompared to LSB steganography. It provides improvement in MSE and PSNRvalues; in addition, the degradation in histogram is also minimal thus eliminatinghistogram attack. Average PSNR value of stego-image obtained from proposedtechnique is 53.11 dB at two bits per pixel data embedding rate compared to 52.21obtained by LSB technique. Keywords  Genetic algorithm (GA)  LSB steganography  Steganalysis P.D. Shah ( & )    R.S. Bichkar G. H. Raisoni College of Engineering and Management, Pune, Maharashtra, Indiae-mail: shahpratik219@gmail.comR.S. Bichkar e-mail: rajankumar.bichkar@raisoni.net  ©  Springer Nature Singapore Pte Ltd. 2018S.S. Dash et al. (eds.),  International Conference on Intelligent Computingand Applications , Advances in Intelligent Systems and Computing 632,https://doi.org/10.1007/978-981-10-5520-1_12119  1 Introduction Steganography is an art of hiding the existence of secret communication [1]. Insteganography, huge amount of secret data is hidden in cover media to conceal it from the attack of eavesdropper  ’ s and unauthorized persons [2]. Steganography canbe done in many digital  󿬁 le formats e.g. video, text, image, audio, etc., but theformats with high degree of redundancy are more suitable. In image steganography,secret data is hidden inside a normal image. The secret information can be in any 󿬁 le format such as text, image, excel  󿬁 le [3]. The image used for hiding the data iscalled as cover image, and the image obtained after embedding the secret data ontothe cover image is called as stego-image. The main aim of steganography is toreduce the difference between stego-image and cover image so that it can concealthe existence of any secret communication. The performance of an imagesteganography technique is mostly evaluated using four parameters viz. imper-ceptibility, payload capacity, robustness and security [1]. Imperceptibility is theability of a steganography technique to be undetected by visual inspection. Payloadcapacity is the amount of secret data that can be hidden inside the cover image.Robustness is the resistance of steganography technique against image manipula-tion attacks like cropping, scaling, rotation, compression. Security is the ability of steganography system to resist the attacks of steganalysis system. Steganalysis isstudy of detecting messages hidden using steganography [4].Image steganography can be categorized as spatial domain steganography andtransform domain steganography. In spatial domain image steganography tech-nique, the data hiding is performed directly on the pixel values of the cover image.Spatial domain techniques include methods which operate at bit level such as bit insertion and noise manipulation [5]. Transform domain techniques utilize thedomain-speci 󿬁 c characteristics of image to embed data on it. The image is  󿬁 rst transformed to frequency domain using numerous transforms like DCT, DFT,DWT, curvelet transform, contourlet transform [6]. In these techniques, the data isembedded on the coef  󿬁 cients of transformed image instead of direct pixels and thenthe image is retransformed to spatial domain.In last decade, enormous amount of work is carried out in the  󿬁 eld of imagesteganography but very few studies have explored the use of metaheuristic andstochastic optimization operators in improving the result of steganography. Kananand Nazeri [7] proposed genetic algorithm-based image steganography technique inwhich they used genetic algorithm to  󿬁 nd the proper locations in cover image tohide secret data. The data was hidden using LSB replacement steganography, GAwas used to  󿬁 nd out starting location and direction for data embedding. The output was tunable, i.e. it generated various different stego-images, any one of which canbe selected based on desired results and application. Average PSNR value of 45.12 dB was obtained during various experiments. Wang et al. [8] proposed asecure steganographic method to bypass RS steganalysis attack. Data is hidden in 󿬁 rst LSB bit of the image, and second LSB bit is modi 󿬁 ed so that image by passesRS attack. Genetic algorithm is used to search for best adjustment matrix which is 120 P.D. Shah and R.S. Bichkar   used to modify second LSB of image so that the stego-image bypasses RS analysisand also provide a better PSNR value. Average PSNR value of 41.2 dB wasobtained during various experiments. Nosrati et al. [9] proposed a before embed-ding steganographic scheme. In this technique, the secret information is hidden inimage segments of cover image. Genetic algorithm is used to  󿬁 nd the suitablelocations in cover image to embed the secret information.The rest of the paper is organized as follows: Sect. 2 brie fl y explains geneticalgorithm. The proposed algorithm is explained in detail in Sect. 3. Experimentalresults and discussion are presented in Sect. 4. Section 5 concludes the paper. 2 Genetic Algorithm Genetic algorithm is biologically motivated metaheuristic technique used to solveboth constrained and unconstrained optimization problems. It is population-basedapproach driven by the principle of natural evolution based on Darwin ’ s theory [10].Each population is potential solution for a given search and optimization problem.Fitness function is used to assess the quality of each possible solution. The solutionwith high  󿬁 tness value will survive and form a new population of the next gener-ation. Genetic algorithm operators like reproduction, crossover and mutation areused to obtain new generations. It is an iterative process which is carried out till thedesired result is obtained or till the number of predetermined iterations is reached.Over successive iterations the population evolves towards a near-optimal solution. 3 Proposed Technique: Optimally Mapped Least Signi 󿬁 cant Bit Replacement Steganography Proposed technique is a modi 󿬁 ed variant of LSB replacement steganography. Inthis technique, secret data is not hidden using LSB replacement method but insteada new approach called data mapping is proposed. In data mapping, two bits of secret data are embedded in every pixel. In grey scale image, each pixel consists of  01000**0 10100**00 2010**100 301**0100 40**00100 5**000100 6*100010* 7Embedding position Coefficient010001** 0 Fig. 1  Co-ef  󿬁 cient generation ( ‘ ** ’  indicates thelocation at which the secret data is embedded)A Secure Spatial Domain Image Steganography  …  121  eight bits, so there are eight possible locations for mapping the secret data bits oneach pixel of cover image. Based on location of match, we generate correspondingcoef  󿬁 cients, as illustrated in Fig. 1.If there is no match between secret data bits and cover image pixel, then thesebits are embedded in LSBs of cover image pixel. It is mostly possible to embed twobits of secret data in each pixel without changing its value. However to successfullyrecover the secret data, we should know the value of coef  󿬁 cients. We propose tohide these coef  󿬁 cients in the unused part of image using genetic algorithm to  󿬁 ndthe most optimal embedding locations.Figure 2 illustrates the proposed mapping process with the help of an example inwhich 4 bytes of secret data is embedded in 4    4 cover image. Figure 2a showsfour bytes of secret data and its binary representation. Figure 2b shows 4    4 cover image and its binary representation along with the position of match between secret data and cover image pixel highlighted. Figure 2c shows the coef  󿬁 cients generatedby mapping.In proposed technique, secret data is embedded in 1/4th part of the image and rest part of the image is used to save the coef  󿬁 cients generated from data mapping. For agrey scale image with 256    256 resolution, we can use 128    128 pixels to hidethe secret data; hence, the resultant data embedding capacity is 2    128    128 =32,768 bits. In this case, a matrix of size 128    128 is generated for the coef  󿬁 cients.From Fig. 2c, it can be observed that value of coef  󿬁 cients is in the range of 0 – 7;hence, each coef  󿬁 cient requires 3 bits for embedding. The coef  󿬁 cients matrix is split into three-bit planes. These three-bit planes are optimally embedded in LSBs of  13 98 57 129 00001101011000100011100110000001 (a) 57 59 62 6871 73 58 9747 56 98 7867 68 98 125 00111 00 1  00 111011 00111 11 0 0100 01 000100 01 11 0100 10 01  00 111010 01 10 0001 00 101111 001 11 000 011000 10  010 01 1100 10 00011 010001 00  0110 00 10 011111 01 (b) 1 6 1 22 2 6 46 3 0 35 0 2 0 (c) Fig. 2  Proposed mapping process to hide secret data in cover image.  a  Secret data and its binaryrepresentation.  b  Cover image (4    4 pixels) and its binary representation.  c  Generated coef  󿬁 cient matrix122 P.D. Shah and R.S. Bichkar 
Search
Similar documents
View more...
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks