You are on page 1of 5


3, AUGUST 2010 141

JPEG Compression Steganography &
Crypography Using Image-Adaptation Technique
Meenu Kumari
BVUCOE/IT Dept, Pune, India

Prof. A. Khare and Pallavi Khare
BVUCOE/IT Dept, Pune, India
SSSIST/E&TC Dept, Bhopal, India

Abstract—In any communication, security is the most technology alone is perfect and can be compromised.
important issue in today’s world. Lots of data security and Once the presence of hidden information is revealed or
data hiding algorithms have been developed in the last even suspected, the purpose of steganography is partly
decade, which worked as motivation for our research. In defeated. The strength of steganography can thus be
this paper, named “JPEG Compression Steganography & amplified by combining it with cryptography.
Cryptography using Image-Adaptation Technique”, we Almost all digital file formats can be used for
have designed a system that will allow an average user to
steganography, but the formats that are more suitable are
securely transfer text messages by hiding them in a digital
image file using the local characteristics within an image.
those with a high degree of redundancy. Redundancy can
This paper is a combination of steganography and be defined as the bits of an object that provide accuracy
encryption algorithms, which provides a strong backbone far greater than necessary for the object’s use and
for its security. The proposed system not only hides large display. The redundant bits of an object are those bits that
volume of data within an image, but also limits the can be altered without the alteration being detected
perceivable distortion that might occur in an image while easily. Image and audio files especially comply with this
processing it. This software has an advantage over other requirement, while research has also uncovered other file
information security software because the hidden text is in formats that can be used for information hiding.
the form of images, which are not obvious text information
Given the proliferation of digital images, especially on
carriers. The paper contains several challenges that make it
interesting to develop. The central task is to research
the Internet, and given the large amount of redundant bits
available steganography and encryption algorithms to pick present in the digital representation of an image, images
the one the offer the best combination of strong encryption, are the most popular cover objects for steganography. In
usability and performance. The main advantage of this the domain of digital images many different image file
project is a simple, powerful and user-friendly GUI that format exit, most of them for specific applications. For
plays a very large role in the success of the application. these different image file formats, different
  steganographic algorithms exist. Among all these file
formats, the JPEG file format is the most popular image
Index Terms—Steganography, Cryptography, Compression, file format on the Internet, because of the small size of
JPEG, DCT, Local Criteria, Image-Adaptation, Huffman the images.
coding, ET, SEC scheme
When working with larger images of greater bit depth,
In simple words, Steganography can be defined as the
the images tend to become too large to transmit over a
art and science of invisible communication. This is
standard Internet connection. In order to display an image
accomplished through hiding information in other
in a reasonable amount of time, techniques must be
information, thus hiding the existence of the
incorporated to reduce the image’s file size. These
communicated information.
techniques make use of mathematical formulas to analyze
Though the concept of steganography and
1 and condense image data, resulting in smaller file sizes.
cryptography are the same, but still steganography
This process is called compression [3]. In images there
differs from cryptography. Cryptography [24] focuses on
are two types of compression: lossy and lossless
keeping the contents of a message secret, steganography
compression [3]. Compression plays a very important
focuses on keeping the existence of a message secret.
role in choosing which steganographic algorithm to use.
Steganography and cryptography are both ways to protect
Lossy compression techniques result in smaller image file
information from unwanted parties but neither
sizes, but it increases the possibility that the embedded
                                                             message may be partly lost due to the fact that excess
Manuscript submitted on May 15, 2010; revised May 17, 2010;
accepted May 31, 2010
image data will be removed. Lossless compression


data embedding is done in the rounded to integer values and the coefficients are transform domain. (These are preserved better not be possible to use with JPEG images. this property can (the decoder guessing that a coefficient has embedded also be used to hide messages. that coefficient. The by grouping the pixels into 8 × 8 pixel blocks and proposed methods can be employed for applications that transforming the pixel blocks into 64 DCT coefficients require high-volume embedding with robustness against each. the Discrete Cosine Transform (DCT) III. data. without any errors. compression would destroy all information in the To compress an image into JPEG format. The results are (a) As is well accepted. changes in the brightness (luminance) of a pixel than to Using the same principles of LSB insertion the message changes in its color. In order to robustly hide large large area. In this representation the Y component lossless stages [3]. the low and mid frequency bands selected as possible Originally it was thought that steganography would candidates for embedding. Thus it is important to recognize that the JPEG colour representation is first converted to a YUV compression algorithm is actually divided into lossy and representation. For JPEG [18]. NO. since the to such a small file size.142 JOURNAL OF ADVANCES IN INFORMATION TECHNOLOGY. In example message would be destroyed. the harsh compression applied. when it actually does not) and deletion errors (the decoder guessing that a coefficient does not have © 2010 ACADEMY PUBLISHER . since they use under compression attacks than high frequency lossy compression [3] which results in parts of the image coefficients) data being altered. although is does not compress the image in an image that uses lossy compression. while the Huffman encoding and V components stand for chrominance (or color). with a set of transform coefficients in encoded using Huffman coding to further reduce the size. it is extremely difficult to detect. This fact is exploited by the JPEG can be embedded into the least significant bits of the compression by down sampling the color data to reduce coefficients before applying the Huffman encoding. (d) The decoder does not have explicit knowledge of During the DCT transformation phase of the compression the locations where data is hidden. used to further compress the data is lossless. the encoder redundant bits of an object and since redundant bits are employs local criteria to select which subset of left out when using JPEG it was feared that the hidden coefficients it will actually embed data in. but not so good as to distinguish between volumes of data in images without causing significant different strengths in high frequency brightness. The next step is the actual transformation of the image. but employs the same algorithm. Even if one could somehow images. the RGB process. A modification of a single DCT coefficient will certain non-malicious attacks. thus domain. VOL. This perceptual degradation. We motivate this by information One of these properties of JPEG is exploited to make theoretic analysis. the use of local criteria for deciding where to keep the message intact it would be difficult to embed the embed is found to be crucial to maintaining image quality message without the changes being noticeable because of under high volume embedding. properties of (c) For each of the selected coefficients. The main diminished. in the transform halved in horizontal and vertical directions. The DCT transforms [18] a signal from successfully. JPEG does this by dividing all the values in a follows. AUGUST 2010 though. propose are guided by the growing literature on the The next step is the quantization [18] phase of the information theory of data hiding [22]. and image tampering. keeps the original digital image intact without the It is neither feasible nor possible to embed information chance of lost. According to research the human eye is more sensitive to Steganography can take place between these two stages. since it is not in decreasing the file size by a factor of 2. 3. the changes to the image invisible to the human eye. rounding errors occur in the coefficient data criteria as the encoder to guess these locations. 1. the data to be the compression algorithm have been exploited in order embedded indexes the choice of a scalar quantizer for to develop a steganographic algorithm for JPEGs. The that are not noticeable. Here another biological property of the The key novelty of our approach is that our coding human eye is exploited: The human eye is fairly good at framework permits the use of local criteria to decide spotting small differences in brightness over a relatively where to embed data. but similar transforms are for example the Discrete Fourier Transform (DFT). without changing the appearance of the ingredients of our embedding methodology are as image. compression. These mathematical We propose a framework for hiding large volumes of transforms convert the pixels in such a way as to give the data in images while incurring minimal perceptual effect of “spreading” the location of the pixel values over degradation. hiding techniques must adapt to means that the strength of higher frequencies can be local characteristics within an image. The DCT and the quantization phase corresponds to the luminance (or brightness) and the U form part of the lossy stage. By the size of the file. decompression. after operations such as an image representation into a frequency representation. However. One of the major characteristics of (b) A novel feature of our method is that. block by a quantization coefficient. The hiding methods we affect all 64 image pixels in that block. additive noise. the visual domain. Although this property is what distortion due to attacks may now lead to insertion errors classifies the algorithm as being lossy. The color components (U and V) are embedding the information at this stage. PROPOSED SYSTEM [18] is used. The embedded data can be recovered part of the image. from the steganography is the fact that information is hidden in the candidate set of transform coefficients.

this joint photographic experts group (JPEG) compression can lead to desynchronization of the encoder and standard. which decides whether or not using simple error correction codes. insertions and deletions are Coefficients (SEC) method. © 2010 ACADEMY PUBLISHER . and that can correct both errors and erasures. can be treated as erasures at the encoder. Both the ET and image tampering attacks. Image-adaptive embedding methodology incur any penalty for erasures at the encoder. However. This We are able to hide high volumes of data under both is because no hidden coefficient with magnitude ≤ t can JPEG and AWGN attacks [24]. assuming that all host coefficients that meet the global criteria will actually be employed for hiding. Hence. that block. AUGUST 2010 143 embedded data. The bit stream to be hidden is coded. Insertions now More advanced image-adaptive schemes would exploit become errors. Moreover. 1.JOURNAL OF ADVANCES IN INFORMATION TECHNOLOGY. where the Two methods for applying local criteria are decoder makes a mistake in correctly decoding the bit considered. It should be noted indicates that the local criteria must be used for robust that a deletion. As seen in the Figure 1. Note that though the PSNR is only marginally attacks. followed by embedding in correcting codes is provided to the synchronization selected DCT coefficients. 3. the hidden be ambiguously decoded to t+1 due to JPEG quantization data also survives wavelet compression. NO. and better. Figure 3. the techniques involve decoder. which causes an erasure. Reed-Solomon codes entropy thresholding (ET) scheme. Turbo-like codes. A maximum distance separable (MDS) code [24]. which decides whether or not observed when the attack quality factor is mismatched to embed data based on the magnitude of the coefficient. In principle. costly as an insertion. such as Reed Solomon (RS) code. image resizing with an interval smaller than the design one. depending on the entropy. hidden). insertions become errors. while a “turbo-like" Repeat for most of the images for JPEG attacks with quantization Accumulate (RA) code is employed for the SEC scheme. and SEC schemes have insertions/deletions under other attacks. with the design quality factor for JPEG attack. and the selectively have been used for ET scheme and Repeat Accumulate embedding in coefficients (SEC) scheme. is about half as and transparent high volume embedding. Reed-Solomon (RS) codes [24] are a natural choice for for the SEC scheme. Both the embedding methods. While the decoding errors can be countered Thresholding (ET) method. it also provides robustness to guessing that there is no data where there was data errors due to attacks. The second is the Selectively Embedding in In the ET scheme. desirable that the data-hiding scheme [4] be adjusted in we refer to our criteria as `perceptual' because our goal in such a manner that there are only a few insertions. one can Figure 1 shows a high-level block diagram of the hiding deal with insertions/deletions without a significant methods presented. taking 2D discrete cosine transform (DCT) of non- (e) An elegant solution based on erasures and errors overlapping 8X8 blocks. A code symbol is erased at the encoder if the local perceptual criterion for the block or coefficient is not met. Thus. There could also be decoding error. using a good erasures and errors correcting code. the actual perceptual quality is much better. VOL. Since we code over entire space of coefficients that lie in a designated low-frequency band. which operate very close to capacity. The subset A threshold is used to determine whether to embed in a of these coefficients in which the encoder does not embed block (ET scheme) or in a coefficient (SEC scheme). using the same local criteria as the encoder). may cause an insertion (decoder guessing that there is While the primary purpose of the code is to solve the hidden data where there is no data) or a deletion (decoder synchronization problem. interval smaller than or equal to the design interval. Coding for Insertions & Deletions: Specifically. are based on codes have been used for the SEC scheme. which causes an error. using local adaptation is to limit perceivable distortion. it is Although we do not use specific perceptual models. does not Figure 1. long codewords can be constructed to achieve very good correction ability. problem caused by the use of local criteria.4 PSNR is better for the image with hidden data using local shows how the sequence is decoded in the presence of criteria. and deletions become erasures (in the human visual system (HVS) models to determine addition to the erasures already guessed correctly by the where to embed information. Coding Framework: The coding framework employs the idea of erasures at the encoder. The first is the block-level Entropy embedded. when it actually does). the decline in original embedding rate. or energy. necessary when hiding large volumes of data into images. incur only a It is observed that the perceptual quality as well as the minor overhead due to erasures at the encoder. we use a code on the hidden data that We noted that use of image-adaptive criteria is spans the entire set of candidate embedding coefficients. using a low rate code. This deletions become additional erasures. As it is seen. there are no insertions or deletions the block-based ET scheme. within synchronization between encoder and decoder. Distortion due to attack decoder. insertions and to embed data in each block (typically 8X8) of transform deletions can potentially cause catastrophic loss of coefficients.

The application is s against statistical primarily intended to be used to inconspicuously hide attacks confidential and proprietary information by anyone Robustnes Low Low Medium High Medium seeking to hide information. making it more difficult to implement. The JPEG file format can be All steganographic algorithms have to comply with a used for most applications of steganography. content of the text message even if its presence were to Unfortunately in the algorithms that are evaluated here. CONCLUSION AND SCOPE FOR FUTURE significant bit (LSB) insertion in BMP and in GIF files. modified. JPEG is the most popular image file information should be hidden in the image.e. To receive a message. The hidden information capacity m should not be lost. Additionally. 3. A combination of the requirement. The main advantage of this ous files program for individuals is that they do not have to have any knowledge about steganography or encryption.) The covering medium Invisibility High* Mediu High High High must be redundant. A short text format on the Internet and the image sizes are small can be used as a key.Depends on cover image used visual way to encode the text. VOL. depending on which requirements are more important for compression is used to maximize the space available in the specific application. AUGUST 2010 IV. encrypted information can be hidden in addition. 128 bit AES encryption is used to protect the algorithm would have a high level in every requirement. for example the cover image used. but only when the most suitable cover image has been Digital Image Steganography [2] system allows a user chosen. re-sized. Independent of file format and Unsuspicious files. sound track. the patchwork approach and spread spectrum techniques. period of time (i. However. The requirements are: especially suitable for images that have to be Invisibility.144 JOURNAL OF ADVANCES IN INFORMATION TECHNOLOGY. TABLE I. The process of embedding information during JPEG To send a message. the application nt of file can also be used by anyone who wants to securely format Unsuspici Low Low High High High transmit private information. Thus a trade-off will exist in most cases. formatted. in other words a perfect. etc. otherwise the hidden information m* could be detected easily. a source because of the compression. (Picture. They can be COMPARISON OF IMAGE STEGANOGRAPHY ALGORITHMS applied together. to securely transfer a text message by hiding it in a digital The ideal. The invisibility. The following table compares least V. if the carrying medium is edited. a source text. no methods are known for there is not one algorithm that satisfies all of the breaking this kind of encryption within a reasonable requirements. while a low level indicates that the Steganography and encryption algorithms provides a algorithm has a weakness in this requirement. 1. A medium strong backbone for its security. according to The meaning of Steganography is hiding information the above requirements: and the related technologies. the process of the © 2010 ACADEMY PUBLISHER .. Robustnes Low Low Medium High High That’s a difficult task to realize. since the embedding takes place in the key is used to aid in encryption and to decide where the transform domain. thus making it the least image containing the information and the corresponding suspicious algorithm to use. but is few basic requirements. Payload capacity. an image in which compression results in a stego image with a high level of the text should be embedded. medium and low. text or even the BMP on structure of a file system. Independe Low Low Low High High Because of its user-friendly interface. There is a principal difference between Steganography and Encryption. manipulation. The * . The technology of hiding should Payload High Mediu Medium Low Medium match the nature of the medium. The levels at which the algorithms satisfy the Digital Image Steganography system allows an requirements are defined as high. plus the visual key makes it easy for average users to navigate within the program. LSB in hiding text in a digital image file or even using it as a key GIF images has the potential of hiding a large message. to the encryption. Currently. RESULT ANALYSIS compression is a very mathematical process. however they can meet at some points too.e. This software has an s against advantage over other information security systems image manipulati because the hidden text are in the form of image. Robustness against image the Internet. Robustness against communicated over an open systems environment like statistical attacks. which is on not obvious text information carriers. NO. compressed or printed. To hide something a covering medium is LSB LSB in JPEG Patch Spread in GIF compressi work spectrum always needed. and a key are needed. steganographic image file. A average user to securely transfer text messages by hiding high level means that the algorithm completely satisfies them in a digital image file. an image. be detected. i. Digital Image level indicates that the requirement depends on outside Steganography system features innovative techniques for influences. a couple of years). WORK JPEG compression steganography.

Anoop K. [8] Fridrich . K.Arce. pp. 2000. Honeyman. [13] M. “The conferences & journals..Ucsb data hiding online demonstration . Joseph Shu.Shamir. the authors and the paper contains the original work of pp. color image secret sharing”. pp. Pallavi Khare.Yang and C.. May 1999. [5] L. Goljan M. 1.ece. C. M. P.A. and Cryptography.” IEEE Security & Privacy "Introduction to Information Theory and Data Journal 2003. “Steganography using higher order Kopargaon. Hankerson.Martens. Codes and Cryptography.Research student of SSSIST. IS&T/SPIE Symposium on Electronic Imaging and Science. [2] N . Persuing M. “ IEEE Trans. of The work on this paper was supported by the Bharati International Conference on Parallel and Distributed Vidyapeeth University & College of Engineering. MA.1913. that doesn't affect the appearance of the image.E. NO.325-335. University College of Engineering. 1-12.2005. Published research paper in one e-journal & one international Proceedings of the International Conference on journals. A. The views and conclusions contained herein are those of [20] R.Tech.” Presented one national comference on Image Compression. Technoogies. The result will appear in the text [16] Daniel L.2. Presented many national & international [10] Mitsugu Iwanmoto and Hirosuke Yamamoto. Systems. papers and [21] R. “The mathematics of halftoning” Farid . Submitted research paper in other national & Information Technology : Coding and computing .International Conference on Digital Printing Proc 10th USENEX Security Symposium 2005. Department.JOURNAL OF ADVANCES IN INFORMATION TECHNOLOGY. Bhopal.1493- 1496. Fundamentals. Res. vol. “Information Theory Coding and Digital Signal Processing Cryptography”. [7] Venkatraman . in   Proceedings of Eurocrypt 1994.Verheul and H.Plantaniotis. Department. [17] C. IBM J. international journals for publication./ data hiding / Currently working as Assistant Professor. Reyzen And S. The Scientist and Engineer's Guide to [24] Ranjan Bose.” uniformity”.950.20.aspx. August 2006. in Proc.C. Inf.R. Irwin Sobel. 2441-2453. [9] http:/ aakash.Lyu and H. 1989 ACKNOWLEDGEMENT [19] C. and M. No. IT from image statistics . from Bhopal. Van Tilborg. 1994. G.Lukac. E&TC [11] Doron Shaked. Bharati Vidyapeeth University College of Engineering.Alder.1. Nur Arad.Arce.1. & Dev.Chang. We took help from many books. Norwood. & Paprzycki M. “Improving void-and-cluster for better halftone [1] N . Pune. October 2002.Lau.Completed B. “Hide and Seek: An [23] D. AUGUST 2010 145 key are both required. s. Khare. and Hogea .Kitchens. HP inherently more difficult to hide information in a JPEG Laboratories Cambridge. 2006. March 2003. India. [4] Katzenbeisser and Petitcolas .P. “Visual Cryptography”. Pune. pp. Russell .8.vol.Zhou. Andrew Fitzhugh. and G. Secur. VOL. 3. Prentice-Hall of India. It is Noise and Green-Noise”. B. “A new approach to other materials. [12] Z.10. Optimal n-out-of-n Visual Secret Sharing Scheme for GrayScale Images”. No. “A new scheme for sharing secret color images in computer network”. vol. ”Information Hiding Techniques for Stenography and Digital watermaking” Artech House. J . Johnson. [6] S.Laih. lecture notes in computer science.47 No. Anil K.E. [18] Jain. and T.. “New colored visual secret sharing image because that is exactly what the designers of JPEG schemes”. On Image Processing. Provos. [15] E. Prof. Significance of Steganography on Data Security “ .15. pp. [3] Steven W. 1997. CA. 1993. Compression”. Smith .Bhattacharjya. 21-27.Smolka. from Sanjivani Educational Society & College of Engineering. Pune. San Jose. Image Systems Laboratory. 2238-2247. A. D . “Fundamentals of Digital Image Processing”. 2000 . “Halftone Visual Cryptography”. in Information Technology Steganography” 2007. and G. Forens. the authors. A . IEICE Trans. HP Laboratories Israel.Chen. Vol. [14] Robert Ulichney. B. Jan.Di Crescenzo. vol. 2003.Naor and A. pp.Ancin. “More efficient provably secure Meenu Kumari. tab after decoding. “ Electronic Imaging 2003”.L. “Color Diffusion: Error Diffusion for Color Halftones”. vol. Harris. Designs. D. Pune University in 2005. Information Technology Released on Mar .332-343. “Fundamental Characteristics of Halftone Textures: Blue- The common Internet-friendly format is offered. “Defending Against Statistical Steganography. pp.R.Completed B. Provos and P. vol. wanted to avoid: the transmission of extra information 2000.Tsai. Designs. “The void-and-cluster method for dither array generation”. EUSIPCO 2004. REFERENCES [22] H. pp. introduction to Steganography. New Methodology for Breaking stenographic Techniques for JPEGs. pp.ucsb. in Bharati Vidyapeeth stegdemo. 2004. Abraham .09.E85- A. IEEE Tans. Gonzalo R. “Constructions and properties of k out of n visual secret sharing scheme”.E. Codes. © 2010 ACADEMY PUBLISHER . Robert Ulichney. 5-15. no.