Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Standard view
Full view
of .
Look up keyword
Like this
0 of .
Results for:
No results containing your search query
P. 1
Neural Networks

Neural Networks

Ratings: (0)|Views: 10 |Likes:
Published by Safaa Shahul Hameed

More info:

Published by: Safaa Shahul Hameed on Oct 26, 2010
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less





  New possibilities of digital imaging and data hiding open wide prospects in modern imagingscience, content management and secure communications. However, despite the obviousadvantages of digital data hiding technologies and their current progress, these developmentscarry inherent risks such as copyright violation, unauthorized prohibited usage anddistribution of digital media, secret communications and network security violations.Although the issues of robustness, visibility and capacity of digital data hiding technologieshave received a lot of attention, their security aspect still remains an open and little studied problem. The security requirement is closely related to the stochastic visibility andunauthorized detection of hidden information and requires both new and careful study. New information-theoretic methods for blind stochastic detection of hidden data should beinvestigated. This aspect will have a great impact on robust digital watermarking,steganography,integrity control and tamper proofing (possibly even without embeddedhidden data) and Internet/network security.
2.Related Research:
An introduction to steganography and its application to digital images are available from[fig5]. Other, more robust, methods of hiding information in images include application of the transform domain that takes advantage of algorithms and coefficients from processing theimage or its components to hide information. Transformations can be applied over the entireimage [8] to blocks through out the image [fig9, fig10], or other variations. Many of thesetransformation techniques require use of the original, unmarked image to extract thewatermark. In [fig11] a number of papers propose techniques that do not require using theoriginal to extract the watermark [fig12]. A method that proposes a combination of thesetechniques from LSB insertion to spread spectrum disbursement of data is described in [fig9].A survey of transform domain techniques can be found in [fig13]. grabs. One commondrawback of virtually all current data embedding methods is the fact that the original image isinevitably distorted due to data embedding itself. This distortion typically cannot be removedcompletely due to quantization, bit-replacement, or truncation at the grayscales 0 and 255[fig6]. Another unconventional approach in which we consider “piggy-backing” the color information on the luminance component of an image for improved color image coding[fig1].The new technique essentially transforms a given color image into the YIQ color space wherethe chrominance information is subsampled and embedded in the wavelet domain of theluminance component
3.The New Approach
 Multiple Key Function Method:
In this paper, we formulate a distinctive method for information hiding in images to beat theexisting detection techniques. Let us take the example of hiding an secret text calledthe
cipher text 
in an image called the
cover image.
 In the effort to keep the structural characteristics of the cover image, secure, we makeminimal introduction of message bits into the cover image and yet manage to embed textmessages of substantial size. It is thus possible, by choosing from a set of key functions,which requires introducing least number of message bits in the cover image and still
accommodate the entire cipher intended to be hidden. The key functions are used to generatea set of byte-positions where the message is to be embedded. At various bit positions in thecover image a match between the cipher bit and the original image bit may result, whichmeans that actually, no replacement of the original image bit occurs. The Bit ReplacementCount (BRC) for this key function is calculated, dynamically, as the replacement/non-replacement of the image bits occur. Here,
BRC = Number of replacements in the original image bits introduced
 Now, another key function from the key function set is chosen and the above manipulationsare performed. As before, the new value of BRC is calculated. Redundant application of theabove formulation will result in a set of BRCs. The set of BRCs can be compared to find outa key function, which belongs to the key function set, which causes least number of originalimage bit replacements. The result is, a sizable text embedded into a cover image, withoutactually disturbing much of the image‘s original characteristics.Associated with the keyfunctions, is a unique identification number in the key functions lookup table. Both the sender and the receiver are required to have a similar lookup table. The unique key identificationnumber is also embedded in the cover image at specific positions in the cover image, alongwith the cipher. The embedded cipher can be recovered safely by the receiver by firstextracting the key identification number from the cover image, and then generate the byte- positions to actually decipher the embedded cipher information. Because the number of bitsin the cover retouched is minimal, as resulted from using the suitable key function, the imagecan escape the statistical analyses (such as CHI-SQUARE or Extended CHI-SQUARE tests),
 Implementation and Testing of Multiple Key method:
This approach was implemented and tested in windows environment with the c++ compiler.The results given above are with a modest, 11 different key function table for a cipher text of 20 bytes. The perfect permutation of the key function and the cover image can produceastonishing results with the ratio of the BRC to the size of the cipher, approaching unity.
 BRC/Size of the Cipher = 1 (approx.)
The results ought to be better if the
key function table
is larger and a suitable database of cover image is maintained. In such cases the probability of BRC would consistently reachunity irrespective of the size of the cipher thus
escaping all existing detection techniques. 
fig 3
3.1.2Enhancements on this technique:
The complexity of the first approach is very high when the number of functions in thefunction domain increases. The complexity can be reduced by using a
generator which dynamic all  generates functions in the neighborhood of the function
that gives better BRC valuesfrom a domain of just 50 functions
3.2 Newton’s Forward Difference Method:
g the above functional table, the forward difference table is constructed. From the tablethus obtained we get the polynomial function that can generate the functional table. This polynomial function is then sent to the receiver along with the image. The receiver canretrieve the polynomial function which is placed in some fixed positions in the image. He canthen generate the functional table thus knowing the bit positions where the cipher information is present. 
In this approach,
 Newton’s forward difference operator 
is introduced. The definition for that operator is
F(x) = F(x+h) - F(x), where h=interval of x values
In the method of Newton Forward Difference, by taking
X :The serial number of cipher bits (1,2,3…n) [ always h=1 ]
F(x): The bit position of image where the respective cipher matches the original Image Bit

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->