Print-Scan Resilient Image Steganography Adriell Dagasuan, student, ADMU, and Adrin Del Rosario, student, ADMU Abstract—This thesis research aims to develop a way of performing a print-scan resilient image steganography with blind decoding . Data would be hidden on an image through a C++ based algorithm. The encoding algorithm will manipulate the image represented in PPMA format. The resulting stega image will then be printed. For the decoding process, the printed stega image will be scanned and the data will be extracted using a C++ based algorithm. Index Terms—steganography, print-scan resilient data hiding, data hiding using yellow markings I. INTRODUCTION TEGANOGRAPHY is the art and science of hiding Smessages; a steganographic system thus embeds hidden data content in unremarkable cover media so as not to arouse an eavesdropper's suspicion. Steganography has been used for hundreds of years and in today’s society there are many uses for steganography. Most modern day uses are in the area of security. Corporate espionage is one way steganography can be used. An employee could get a job at a specific company with the intention of stealing valuable information from them. One way to pass the information back would be through steganography. In today’s society the most practical implementation of steganography is used in the world of computers. Data is the heart of computer communication and over the year a lot of methods have been created to accomplish the goal of using steganography to hide data. The best object up to this writing is probably a digital image. Digital images have the benefit of containing massive amounts of bytes to designate pixel color for the photo. II. SIGNIFICANCE OF THE STUDY This thesis research aims to extend the range of applications of steganography to printed images by developing an algorithm that allows a print-scan resilient steganography. III. SUMMARY OF PREVIOUS RESEARCHES DONE A. Fujitsu's Implementation Fujitsu’s technique works by taking advantage of the sensitivities of the human eye, which struggles to see the colour yellow. The key is to take the yellow hue in the picture skew it slightly to create a pattern. A camera is perfectly sensitive to that yellow hue but the human eye doesn’t see it very well. The implementation divides the original image into smaller blocks of 0.8mm square or less. The average gradation or density level of each block is analyzed and then the information is added as a sequence of yellow dots which have a lower gradation. Depending on the bit value, the implementation would place the yellow dots on the left side or right side of a cell. B. Print and Scan Resilient data Hiding in Images Solanki, et. al, proposed methods to hide information into images that achieve robustness against printing and scanning with blind decoding. The selective embedding in low frequencies scheme hides information in the magnitude of selected low-frequency discrete Fourier transform coefficients. The differential quantization index modulation scheme embeds information in the phase spectrum of images by quantizing the difference in phase of adjacent frequency locations. A significant contribution of their paper is analytical and experimental modeling of the print-scan process, which forms the basis of the proposed embedding schemes. A novel approach for estimating the rotation undergone by the image during the scanning process is also proposed, which specifically exploits the knowledge of the digital halftoning scheme employed by the printer. Using the proposed methods, several hundred information bits can be embedded into images with perfect recovery against the print-scan operation. IV. SIGNIFICANT RESULTS A. Images undergoing the printing and scanning process We tried to see how much an image differs once it has been printed and scanned (print-scan image). We used The GIMP to obtain the PPMA format of the images. We tried comparing various images from its print-scan counterpart and the results were all the same. The image changes significantly after it has undergone the printing and scanning process. We took this heavily into our consideration in planning for our encoding and decoding algorithm. B. Using pure yellow pixels We tried to implement the steganography implementation used by Fujitsu. To hide data, we tried replacing some pixels with pure hues of yellow (R=255, G=255, B=0). Our algorithm asks for a pattern. This pattern determines which pixels in an image cell would be replaced by yellow pixels. Although we could specify a pattern similar to Fujitsu's,we specified other patterns that could probably reduce the size of the cell, thus increasing the data capacity of an image. Once the pattern, the name of the cover image, and the data have been specified, the encoding process will start. This will replace the pixels, specified by the pattern, with yellow pixels regardless of the original values of these pixels. The resulting stega image was very much degraded. It won't pass as a stega image because the yellow pixels are very obvious and significantly changed the image. The printed image is the same. Suspicious-looking yellow dots can be easily identified. This result prompted us to change our approach in changing the values of the pixels. D. Decoding the pre-print-scan stega image C. Using a factor to bring out the yellow Since giving the pattern-specified pixels with pure yellow hues gives away the presence of data, we tried another approach. This time, we considered the original values of the pixels; Instead of replacing the pixels with pure yellow hues (R=255, G=255 ,B=0), we reduced the blue component to a certain factor. For example, we multiplied the blue component of the pixel with 0.5. This gives the pixel a yellow tint. The yellow dots becomes more obvious as the factor is reduced. Using a factor of 0.1, for example, makes the yellow dots in the stega image more visible as compared to a stega image which used a factor of 0.8. The resulting stega images are satisfactory until about a factor of 0.5. We digressed from the Fujitsu implementation by using a single pattern but with different factors to differentiate the 0-bit from the 1-bit. We arbitrarily assigned a factor of 0.8 to 0-bit and a factor of 0.5 to 1-bit. For the decoding process, we analyzed a cell to determine if it is embedded with 0-bit or 1-bit by getting the ratio of the values of the blue component of pattern-specified pixels (those whose blue component values were reduced) and pattern-unspecified pixels (those whose blue component values were retained). This ratio corresponds to the factor that was used in the encoding process. By changing certain parameters in the algorithm, we were able to achieve up to 100% data recovery. E. Decoding the print-scan stega image Although the decoding process worked well for the preprint-scan stega image, it did not work out at all for the printscan image. Upon inspection, the ratio between the patternspecified and pattern-unspecified pixels differed greatly from the factor used during the encoding process. This is due to the printing and scanning process. Our next step is to assimilate in our encoding process, an image analysis that would verify if the data is recoverable from the print-scan image REFERENCES [1] [2] [3] Fujitsu Ltd., “Printed Steganography Embedding Invisible Data in a Picture Image,”Science Links Japan, 2005. K. Solanki, U. Madhow, B. S. Manjunath, S. Chandrasekaran, and I.ElKhalil, “Print and Scan Resilient data Hiding in Images,” IEEE Trans . Information forensics and security, vol. 1, no. 4, December 2006. M. S. Fu and O. C. Au, “Data hiding watermarking in halftone images,” IEEE Trans. Image Process., vol. 11, no. 4, April 2002. Adriell Matthew Julius A. Dagasuan Adrin V. Del Rosario