Ndd

New Dimension of Data Security using Neural Networks and Numerical Functions

TABLE OF CONTENTS 1.ABSTRACT OF THE PAPER 2. INTRODUCTION TO DATA HIDING 3. OTHER RELATED WORK 4.THE NEW APPROACH 4.1 MULTIPLE KEY FUNCTION METHOD 4.1.1 IMPLEMENTATION AND TESTING 4.1.2 ENHANCEMENTS ON THIS METHOD 4.2 NEWTON’S FORWARD DIFFERENCE METHOD 4.2.1 ADVANTAGES OF THIS METHOD 4.2.2 IMPLEMENTATION AND TESTING 4.2.3 ENHANCEMENTS ON THIS METHOD 4.3 COMPARISON BETWEEN THE EXISTING TECHNIQUES AND OUR TECHNIQUE.

5.CONCLUSION

6.REFERENCES

New Dimension of Data Security using Neural Networks and Numerical Functions

1.Abstract:

Data hiding, embeds data into digital media for the purpose of identification, annotation, and copyright. New possibilities of digital imaging and data hiding open wide prospects in modern imaging science, content management and secure communications. Our objective is to hide large volumes of data in the host, in a manner that causes minimal perceptual distortion, and is robust to survive benign and malicious attacks, and make sure to hide as much information as possible into the host. In this paper we identify the prospects of the hidden information being detected and introduce a new approach for hiding any text without any embedding, by the method of Multiple key functions and Newton Forward Difference Technique so that it passes unnoticed by the

existing stochastic detection techniques. By this approach, the size of the data that can be hidden will be larger than all the size that all the existing techniques support. This approach derives a polynomial function that generates the bit position in the host where the data bits matches with the host bits. Since the approach uses the original bits of the host to hide the data, the conventional embedding is not done and so the host does not undergo any perceptual degradation and thus escapes all the existing stochastic detection techniques.

2. Introduction :

New possibilities of digital imaging and data hiding open wide prospects in modern imaging science, content management and secure communications. However, despite the obvious advantages of digital data hiding technologies and their current progress, these developments carry inherent risks such as copyright violation, unauthorized prohibited usage and distribution of digital media, secret communications and network security violations. Although the issues of robustness, visibility and capacity of digital data hiding technologies have 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 and unauthorized detection of hidden information and requires both new and careful study. New information-theoretic methods for blind stochastic detection of hidden data should be investigated. This aspect will have a great impact on robust digital watermarking, steganography, integrity control and tamper proofing (possibly even without embedded hidden data) and Internet/network security.

3. Related Work:

An introduction to steganography and its application to digital images are available from [5]. Other, more robust, methods of hiding information in images include application of the transform domain that takes advantage of algorithms and coefficients from processing the image or its components to hide information. Transformations can be applied over the entire image [8] to blocks through out the image [9, 10], or other variations. Many of these transformation techniques require use of the original, unmarked image to extract the watermark. In [11] a number of papers propose techniques that do not require using the original to extract the watermark [12]. A method that proposes a combination of these techniques from LSB insertion to spread spectrum disbursement of data is described in [9]. A survey of transform domain techniques can be found in [13]. grabs. One common drawback of virtually all current data embedding methods is the

fact that the original image is inevitably distorted due to data embedding itself. This distortion typically cannot be removed completely due to quantization, bit-replacement, or truncation at the grayscales 0 and 255 [6]. Another unconventional approach in which we consider “piggy-backing” the color information on the luminance component of an image for improved color image coding[1]. The new technique essentially transforms a given color image into the YIQ color space where the chrominance information is subsampled and embedded in the wavelet domain of the luminance component

4. The New Approach: 4.1 Multiple Key Function Method:

In this paper, we formulate a distinctive method for information hiding in images to beat the existing detection techniques. Let us take the example of hiding an secret text called the cipher text in an image called the cover image. In the effort to keep the structural characteristics of the cover image, secure, we make minimal introduction of message bits into the cover image and yet manage to embed text messages 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 generate a set of byte-positions where the message is to be embedded. At various bit positions in the cover image a match between the cipher bit and the original image bit may result, which means that actually, no replacement of the original image bit occurs. The Bit Replacement Count (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 manipulations are performed. As before, the new value of BRC is calculated. Redundant application of the above formulation will result in a set of BRCs. The set of BRCs can be compared to find out a key function, which belongs to the key function set, which causes least number of original image bit replacements. The result is, a sizable text embedded into a cover image, without actually disturbing much of the image‘s original characteristics.

Associated with the key functions, 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 identification number is also embedded in the cover image at specific positions in the cover image, along with the cipher. The embedded cipher can be recovered safely by the receiver by first extracting 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 bits in the cover retouched is minimal, as resulted from using the suitable key function, the image can escape the statistical analyses (such as CHI-SQUARE or Extended CHI-SQUARE tests),if the cover image is carefully chosen.

.

4.1.1 Implementation and Testing: 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 produce astonishing 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 reach unity irrespective of the size of the cipher thus escaping all existing detection techniques.

4.1.2 Enhancements on this technique: The complexity of the first approach is very high when the number of functions in the function domain increases. The complexity can be reduced by using a generator which dynamically generates functions in the neighborhood of the function that gives better BRC values from a domain of just 50 functions.

4.2 Newton’s Forward Difference Method:

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

Using the above functional table, the forward difference table is constructed. From the table thus 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 can retrieve the polynomial function which is placed in some fixed positions in the image. He can then generate the functional table thus knowing the bit positions where the cipher information is present.

4.2.1 Advantages of this method:

Since the method uses the original bits of the image to hide the cipher, the conventional embedding is being ruled out and so probability of BRC is consistently equal to 1 for any cipher size and any cover image. Since we are not going to choose all the bits of the image in sequence, the size of the cipher that can be hidden is extremely large almost equal to half of the size of the image. Since we are not going to embed, the statistical properties of the image left unchanged. So the image escapes all the statistical detection techniques.

4.2.2 Implementation and testing: This new approach was implemented and tested in windows environment with the c++ compiler. The forward difference table was constructed for a cipher length of 10 bytes., that is, 80 bits and the functional table obtained is as given below.

The above functional table yields the polynomial function 2x^5+2x^4-3x^2+2 after constructing the forward difference table. The technique was tested for five cases by hiding different ciphers in various images and the results thus obtained are shown in the graph.

Thus this approach proves to be more efficient than any other method because nowhere we are going to replace or embed any data into the image. Any detention technique (CHI-SQUARE method) fails to detect any suspection over the image.

4.2.3 Enhancements on this technique:

The degree of the polynomial thus obtained is very high .To reduce the degree of the resulting Polynomial function, we use a selector to which some intelligence is incorporated. We use the concepts of neural networks to achieve this task. We try to test and train the selector so that it intelligent selector such that the image bit positions such that the higher order differences in the forward difference table vanishes to give a polynomial of lesser degree.

As shown above, the same technique is applied recursively on the resultant polynomial functions Until the resultant polynomial function is of lesser degree. The polynomial table resulting because The enhancements is shown below. This table corresponds to the polynomial 3x-2.

4.3 Comparison between the exitsting techniques and our method:

5. Conclusion :

With the arrival of the digital society in the 21st century, the potential of information hiding and watermarking techniques for multimedia has drawn increasingly intensive attention from the research communities and commercial sectors all over the world during the last decade. Recently, many researchers are investigating image data hiding techniques in various aspects in addition to imperceptibility and robustness. These aspects include information theoretic, stochastic, and security issues. This research trend will advance the information hiding technologies and provide opportunity for practical systems. This paper proposes

two new evolutionary techniques in the field of data hiding. The first approach aims at minimal embedding by increasing the number of functions. This approach provides facilities for using 65536 functions as we have used a 16 bit key. The second approach aims at avoiding embedding thus satisfying theoretic, stochastic, and security issues.

6. References

[1] Compressive data hiding: an unconventional approach for improved color image coding by patrizio campisi, deepa kundur, dimitrios hatzinakos, and alessandro neri

[2] Neil F. Johnson and Sushil Jajodia, “Steganalysis: The Investigation of Hidden Information”, Proceedings of the 1998 IEEE Information Technology Conference, Syracuse, New York, USA, September 1st - 3rd, 1998.

[3] N.F. Johnson, S. Jajodia, "Steganalysis of Images Created Using Current Steganography Software", Proceedings of Information Hiding Workshop, Portland, Oregon, USA, April 1998.

[4] Niels Provos, “Probabilistic methods for improving Information hiding”, Center for Information Technology Integration – Technical Report 01-1, January 31, 2001.

[5] N.F. Johnson, S. Jajodia, "Exploring Steganography: Seeing the Unseen", IEEE Computer, February 1998, vol. 31, no. 2, pp.26-34

[6 ] “ Lossless data embedding” —new paradigm in digital watermarking by jessica fridrich, miroslav goljan, and rui du.

[7] T.G. Handel, M.T. Stanford, III. "Hiding Data in the OSI Network Model", In: [7] pp. 23-38, 1996.

[8] I. Cox, J. Kilian, T. Shamoon, T. Leighton, "A Secure, Robust Watermark for Multimedia", In: [1] pp 185- 206, 1996.

[9] G.B. Rhoads, "Method and Apparatus Responsive to a Code Signal Conveyed Through a Graphic Image", U.S. Patent 5,710,834, January 20, 1998. Held by Digimarc Corporation, http://www.digimarc.com

[10] M.D. Swanson, B. Zhu, A.H. Tewfik, "Transparent robust image watermarking", Proceedings of IEEE International Conference on Image Processing (ICIP96), Piscataway, NJ. IEEE Press, vol. III, 1996, pp. 211-214.

[11] Proceedings of .IEEE International Conference on Image Processing (ICIP'97), Santa Barbara, California, October 26-29, IEEE Press, 1997. [12] A. Piva, M. Barni, F. Bartolini, V. Cappellini, "DCTBased Watermark Recovering Without Resorting to the Uncorrupted Original Image", in [15], 1997 pp. 520-523. [13] N.F. Johnson, Z. Duric, S. Jajodia, "The Role of Digital Watermarking in Electronic Commerce", submitted for publication to ACM issue on electronic commerce. [14] E. Franz, A. Jerichow, S. Moller, A. Pfitzmann, I.Stierand, "Computer Based Steganography: How it works and why therefore any restrictions on cryptography are nonsense, at best", In: [1] pp. 7-21, 1996.