[7] Ekta Walia, Payal Jain, Navdeep, âAn Analysys of LSB & DCT based ... [8] Siddharth Singh, Tanveer J. Siddiqui, â A Security Enhanced Robust Robust ...
Proc. of Int. Conf. on Advances in Computer Science, AETACS
Generic Function Based Color Image Steganography in 2-D DCT Domain Rajib Biswas a , Rajarshi Basak b, Samir Kumar Bandyopadhyay c a
b
Department of Information Technology, Heritage Institute of Technology, Kolkata-107, India. Department of applied Electronics and Instrumentation, Heritage Institute of Technology, Kolkata-107. India. c Department of Computer Science and Engineering , Calcutta University, Kolkata-9, India.
Abstract— Steganography is a means of establishing secret communication through insecure public channel in an artistic manner. In this paper, we propose a generic function based steganography in DCT domain, applying 2-D DCT in colour images(GFBSD). The cover image is first transformed through a process , then segmented into 8X8 blocks after applying a mathematical function to make it more secured and DCT is applied on each block. Now after processing the secret data, a CONCEALER is applied on it functionally and secret function (secret data with functional concealer) is embedded into middle bits of DCT coefficients of high frequency range. It has been observed that the proposed algorithm has better PSNR, MSE, security and data hiding capacity compared to the existing algorithms. Index Terms— Steganography, Steganalysis, Cover Image, DCT Transform, Inverse DCT Transform, Concealer.
I. INTRODUCTION In present electronic communication scenario, data security is one of the major challenges. After World WarII, the need of a secured and robust communication between the entities has increased due to the fear of terrorism. The publisher of digital audio and video are worried of their works being corrupted by illegal copying or redistribution. Hence it is very much important to protect it. Cryptography, from the ancient past is a method of hiding secret data by scrambling so that it is unreadable, though, it does not assure security and robustness as the unauthorized person can obviously guess that there is a confidential message passing on from source to the receiver. Steganography[4] is concealed writing and it is a scientific approach of inserting secret data within a cover media and to do it sophistically so that unauthorized viewer do not get idea of any communication happening. Steganography with cryptography provide high level security to the secret information. Cover image is known for carrier image and it is the original image in which the secret data is embedded. II. LITERATURE STUDY An extensive study of the related papers [2],[3],[5],[6],[7],[8] has given shape to this concept. The authors have meticulously analyzed the possibilities in the sphere of maximizing randomization , minimizing deviations, increasing integration [4] and structuring strong coherence among the working sets. This paper is aimed at further increasing the equalization and reliability of the DCT based color image steganography[11] © Elsevier, 2013
from its referrals. III. DISCRETE COSINE TRANSFORM The discrete cosine transform is quite closely related to DFT(Discrete Fourier Transform) but offers high energy compaction property in comparison with DFT for natural images. It’s a real domain transform which represents an image as coefficients of different frequency of cosine which is basis vector for this transform[1]. The general equation for 2-D (NXM data items) DCT is defined by the following:
Similarly the Inverse DCT can be calculated using the following equation, where the signal comes back to spatial domain [10] from Transform domain.
IV. OUR PROPOSED METHOD The proposed method comprises mainly two components namely embedding process and retrieval process. The embedding process consists of cover image, pixel management, color code function, 8×8 block segmentation[11], DCT, secret function formation, encryption, IDCT, inverse color code formation, stego image formation. The recovery process is again divided into stego image, pixel management, color code function, 8×8 block segmentation, DCT , retrieval of by parted secret function, secret data recovery by applying inverse of generic function. A. Embedding process The secret data in the form of secret function after using concealer is embedded according to the process described. 1. Cover Image: The cover image in Fig.1. can be color or gray scale image of any size and format. 2. Pixel management: As we are applying 2D-DCT on the color image, a color image is managed by a pixel management function. Embedding image = f1 (cover image).
Fig. 1. Cover Image.Lena(405x402)
Fig. 2. Histogram of Original Image.
3. Color code Function: To make the algorithm one level secured, the pixel values of the embedding image are varied by the color code function. Embedding space= f2 (embedding image).
4. 8×8 segmentation: The embedding space is segmented into 8×8 matrices for applying DCT. 5. Applying DCT: Transform each 8×8 matrix into frequency domain using 2D-DCT “(1),”. Frequency domain Embedding space =DCT (E.S). 6. Secret Function Formation: The secret data is taken from the user and it is functionally modified into secret function by applying concealer which is a varying function and user can change it according to the requirement. Here f3 is a generic mathematical function. Secret function = f3(secret data, concealer). 7. Embedding or encryption: The secret function is now divided into two parts. This is done deliberately to make the distortion in the stego Image as less as possible along with the increasing security of the algorithm . Then the first part/MSBs and LSBs are encrypted separately in the DCT matrix with suitable DCT coefficients in the middle bit position. 8. IDCT: After embedding the secret function the IDCT is done on each 8×8 matrix following the equation 2. Embedded space=IDCT(f.d. Embedded space). 9. Inverse color code function: the inverse color code function is done on the embedded space to get back the embedding image. Embedding image=f2-1(Embedding space). 10. Stego Image Formation: The stego image in Fig. 3. is formed from the embedding image pixel values. 4.2. Retrieval process The stego image being received at the receiver end, the secret data is retrieved through the following process. 1. Cover Image: The cover image can be color or gray scale image of any size and format. 2. Pixel management: As we are applying 2D-DCT on the color image, a color image is managed by a pixel management function. Embedding image = f1 (cover image). 3. Color code Function: To make the algorithm one level secured, the pixel values of the embedding image are varied by the color code function. Embedding space= f2 (embedding image). 4. 8×8 segmentation: The embedding space is segmented into 8×8 matrices for applying DCT. 5. Applying DCT: Transform each 8×8 matrix into frequency domain using 2D-DCT.(Equation 1) Frequency domain Embedding space =DCT (E.S). 6. Recovery of by parted secret function: The system function is recovered from the frequency domain embedding space part by part. 7. Secret function formation: The secret function is obtained from the by parted secret function . 8. Secret data recovery by applying inverse of generic function : The secret data is recovered by the receiver by applying the concealer with inverse generic function. The concealer should be provided to the receiver end as without it the retrieval is not possible. Secret data= f3-1(secret function, concealer).
Fig. 3. Stego Image Lena(405x402).
Fig. 4. Histogram of Stego Image.
V. PERFORMANCE ANALYSIS Steganolysis: Through steganolysis is done to check the immunity of the algorithm [10] and to clarify the proximity and negligible changes produced in the stego image. Two images above shown are the Original
image of ‘Lena’ and its stego form after the encryption is done. Along with that, the histograms of the R,G,B values of the original image in Fig. 2. and the stego image Fig. 4. are given side by side. From the table underneath, Table 1, it is noted that the statistical parameters like the MEAN, STANDARD DEVIATION and VARIANCE change only in their distant decimals thus proving its strong resistance to steganalysis [12]. Some more analysis has also been done to emphasize the minute or almost negligible deviation of the stego image from the original image. MSE: It is defined as the square of error between original image and the stego image and the MSE test results is shown in Table 2 .
f(x,y)= The intensity of the pixel in the original image. g(x,y)= The intensity of the pixel in the stego image. TABLE I - MEAN , VARIANCE, STANDARD DEVIATION Statistical Parameters
Image
MEAN VARIANCE SANDARD DEVIATION
Original 71.3067 4.3272e+003 65.7814
Stego 71.3057 4.3272e+003 65.7818
TABLE II – MSE RESULT MSE Color Component Result
R 4.8678
G 4.9147
B 5.0655
TABLE III – STIRM ARK ANALYSIS OF E MBEDDING ALGORITHM ON LENA IMAGE (405 X 402 SIZE) OF FIG. 1 AND FIG. 3 Test SelfSimilarities SelfSimilarities SmallRandomDistortions SmallRandomDistortions MedianCut MedianCut PSNR PSNR AddNoise AddNoise
Factor 1 2 1.05 1.1 3 9 10 20 20 40
Cover 28.529 dB 44.807 dB 13.846 dB 13.724 dB 30.857 dB 22.197 dB 38.662 dB 34.236 dB 9.432 dB 7.879 dB
Steg 28.519 dB 44.807 dB 11.675 dB 11.391 dB 29.478 dB 22.219 dB 38.651 dB 34.224 dB 9.441 dB 7.886 dB
StirMark Analysis: Any steganographic algorithm should resist some standard benchmark tests to prove its strength and robustness. We run these tests in StirMark 4.0 [9] and our algorithm produced good results. We show a sample of the results in Table 3. The negligible gaps between the values corresponding to the cover and the stego image imply that our technique is robust. PSNR(Peak signal Noise Ratio): It is the measure of the quality of the image by comparing the original image in Fig.1. with the stego image in Fig.3., i.e. it measures the statistical difference between the original and the stego image. PSNR=10*log(2552/MSE). VI. FUTURE ENHANCEMENT We would like to step forward with steganography into the field of audio-visual files with transform methods.
VII. CONCLUSION In this paper, the authors have proposed a steganography process with a colour image in discrete cosine transform (DCT) domain to improve security and stego image quality compared to the existing algorithms which are normally done in spatial domain. According to the simulation results the RGB values of stego images of our method are almost identical to original images and it is very difficult to differentiate between them visually. Our proposed algorithm also provides additional two layers of security by means of transformation (DCT and Inverse DCT) of cover image. Hence the proposed method may be more robust against brute force attack. That means secret image keep the secret massage away from stealing, destroying from unintended users. ACKNOWLEDGEMENT We express our gratitude to all those who have in some way or the other been a part of this enterprise REFERENCES [1] Digital Signal Processing-System Analysis & Design by Paulo S.R.Diniz, Eduardo A.B. da Silva, Sergio L.Netto. Cambridge University Press. ISBN No. 9780521887755,September 2010. [2] J.R.Krenn, “Steganography and Steganolysis”,January 2004. [3] Hsien – Wen Tseng and Chin – Chen Chang, “High Capacity Data Hiding in Jpeg Compressed Images”, Informatica Volume-15, Issue 1(January 2004) 127-142,2004,0868-4952. [4] Youngran Park, Hyunho Kang, Kazuhiko Yamaguchi and Kingo Kobayashi, “Integrity Verification of Secret Information in Image Steganography”, Symposiun on information Theory and its Applications, Hakodate, Hokkaido, Japan, 2006. [5] KokSheik Wong, Xiaojun Qi and Kiyoshi Tanaka, “ A DCT Based MoD4 Steganographic Method”, Signal Processing 87, 1251-1263,2007. [6] Takayuki Ishida, Kazumi Yamawaki, Hideki Noda, Michiharu Niimi, “Performance Improvement of JPEG Steganography Using QIM”, Department of System Design and Informatics, Journal of Communication and Computer, ISSN1548-7709, USA, Volume 6, No. 1(Serial No.-50), January 2009. [7] Ekta Walia, Payal Jain, Navdeep, “An Analysys of LSB & DCT based Steganography”, Global Journal of Computer Science and Technology,Volume 10, Issue 1(Ver 1.0), April 2010. [8] Siddharth Singh, Tanveer J. Siddiqui, “ A Security Enhanced Robust Robust Steganography Algorithm for Data Hiding”, International Journal of Computer Science Issues, Volume 9, Issue 3, No. 1,May 2012. [9] Fabien A. P. Petitcolas, Ross J. Anderson, Markus G. Kuhn. Attacks on copyright marking systems, in David Aucsmith (Ed), Information Hiding, Second International Workshop, IH’98, Portland, Oregon, U.S.A., April 15-17, 1998, Proceedings, LNCS 1525, Springer-Verlag, ISBN 3-540-65386-4, pp. 219-239. [10] Paul, G., Davidson, I., Mukherjee, I., Ravi, S.S.: Keyless Steganography in Spatial Domain using Energetic Pixels. In: Venkatakrishnan, V., Goswami, D. (eds.) ICISS 2012. LNCS, vol. 7671, pp. 134–148. Springer, Heidelberg (2012). [11] “DCT Domain Encryption in LSB Steganography”, Biswas, R., Mukherjee, S. ; Bandyopadhyay, S.K. CICN, 2729, Sept 2013. [12] Rajib Biswas, Gaurav Dutta Chowdhury, Samir Kumar Bandhyopadhyay “Trigonometric Abstraction Based Variable Harmonic Sampling and Dual Encryption Steganography” International Journal of Scientific & Engineering Research ,ISSN 2229-5518,,Volume 4, Issue 7, July 2013.