Computerized Image Processing for Forgery Prevention Nathan Singleton Computing Research Department of Computing Sciences Villanova University, Villanova, PA, 19085 USA nathan.singleton@villanova.edu ABSTRACT The issue of copyright protection for digital media has become a problem with rapid advancements in online media storage. As more of these storage systems are made, more information will be taken from them. For all of the good that comes from new technology, an equal amount of illegal activity can spore from it. Digital watermarking has become one of the most popular copyright protection methods. This paper presents different ways in which watermarking is used, along with the background of watermarking for personal protection in general. There are many different aspects of digital watermarking. This paper focuses on the main algorithms used in the present day. It will also look into the future of digital watermarking and how it will be able to provide more reliable copyright protection. KEY WORDS Digital Watermarking, Fragile, Robust, Visibility 1. INTRODUCTION Digital Watermarking is a process that embeds data, called a watermark, digital signature, or tag, into a multimedia object. The watermark can then be extorted or detected to make a claim about the multimedia object [1]. The multimedia object being watermarked can take many forms, such as a digital picture, audio file, or video clip. Watermarks have a variety of uses that include personal protection and deterrence against theft to maintaining data integrity, secrecy and security. Multimedia that is stored on computers, and specifically in the case of the example pictures included in this paper, has always been easy to copy between computers with minimal technical expertise. The use of computers to transfer, post, and share work is also a relatively simple matter by the means of websites and peer-to-peer file sharing systems. With this straightforward and widely available ability to share information comes an increasingly onerous hardship for rightful owners of digital media. Among the most common purposes of watermarking are to maintain the rightful ownership of digital media by the copyright holders, prevent or discourage the pirating of digital media, and ensure that digital media is not resold without permission of the media owner. 2. BACKGROUND The first use of watermarking of personal property for purposes of security or protection is a matter of continuing debate, although evidence indicates it may have occurred in Bologna, Italy in 1282 [2]. This presumed first watermark was found on an Italian piece of paper from that time. It represented a “pomme cross with small circles at the ends of the arms of the cross and a large one at the centre”[2]., and although the purpose of it is still under debate as well, there is reason to believe that watermarks were used for similar reasons as they are used today. Some even speculate that certain watermarks “were used by mystical groups to carry signals with hidden meanings,” somewhat similar to what they are used for today [2]. Today, watermarks can be visible, such as those embedded into images or video and designed to be seen by the human eye, or invisible, only detectable by computer software. Figure 1: First know watermark [3] 3. WATERMARKING The general form of all algorithms used to input a watermark in an image involves the original watermark, and the original image. The only main step found in all watermarking techniques is the Embedding Procedure, which places the watermark in the image. The last step is the output, or the watermarked image. This can be seen in figure 1. The purpose of invisible watermarks is similar to the purpose of visible watermarks, but with a level of secrecy instilled. Content identification is one reason for digital watermarking, because the rightful owner of the product wants to have the security to know the image is his. Copy protection is another reason for invisible watermarking that gives the rightful owner the ability to disallow people from using illegal copies of the image for commercial use. Viewer tracking is a more technical use for invisible watermarking that is especially useful for images posted on the internet. Viewer tracking allows the owner to track how often the image is viewed. The owner of the image, with advanced software, can also detect if his image has been posted illegally on other websites. 3.2 ROBUSTNESS Figure 2: Watermarking algorithm general form [4] 3.1 VISIBILITY Today all watermarking algorithms can be broken down into two main catagories, visible watermarks and invisible watermarks. Visible watermarking techniques are generally easily implemented and are found and can be used in many, if not all photo editing software, such as Adobe Photoshop or even Microsoft Paint. However if these watermarking techniques are used the original image might be damaged or destroyed completely. “Visible watermarks are designed to be easily perceived by the view, and clearly identify the owner; the watermark must not detract from the image content itself, however. Most research currently focuses on invisible watermarks, which are imperceptible under normal viewing conditions” [3]. Visible watermarks should be used in digital libraries so that they can be used in research papers and other scholarly works. This keeps the image safe from commercial use. Figure 3: Example of Visible Watermark [5] There are two different types of robustness that are generally taken into account when evaluating a watermarking algorithm, Invisible –Fragile Watermarking and Invisible – Robust Watermarking. “Invisible –Fragile Watermarking is embedded in such a way that any manipulation or modification of the image would alter or destroy the watermark” [1]. In other words, the watermark in the image would be destroyed if simple modifications to the image are done. Fragile watermarks are used in order to detect tampering. This is the reverse of how one would think a watermark should work. If the picture is tampered with the watermark will disappear and be undetectable, and therefore the culprit will be caught. An attacker of a fragile watermark would be successful if he was able to alter the picture without disturbing the watermark. “Fragile watermarks are used for image authentication. Image authentication systems have applicability in law, commerce, defense, and journalism” [4]. A common example is the marking of images in a memory card inside a camera. This would ensure that photographers provide trustworthy news to consumers, and do not broadcast false events. In most cases the watermark placed into a fragile watered image cannot be taken out. For example when this implementation is used in a digital camera, the watermark is placed directly after the image is taken, and the owner does not know the difference. Host Pixel: 10110001 Secret Pixel: 00111111 New Image Pixel: 10110011 Figure 5: Least Significant Bit Algorithm example [8] Figure 4: General Algorithm for Fragile Watermark Detection [6] Robust Watermarking has proven to be a more prevalent research topic in the area of Digital Watermarking. “Robust watermarks withstand moderate to severe signal processing attacks (compression, rescaling, etc.) on an image” [4]. In other words, they have the ability to be compress, such as with jpeg compression, and keep the information stored inside. Robust Watermarks also keep their information hidden when cropped or rotated as well, unlike with all fragile watermarks. A robust watermark is implemented by embedding a host signal in an image so that the removal of the said signal is complex without corrupting the original image [5]. This type of invisible watermarking has many real world applications. It is used, for example, when the seller of digital images in concerned that his fee-generating images may be obtained by a person who will make them available for free [1]. This would make it so that the owner is not paid for the work put into the product. This is seen as a fragile technique, because if the new image is tampered with it is easy for computer software to detect. This is because the software that would detect these interferences only has to compare it with an originally made image. A different watermarking algorithm involves Discrete Cosine Transform or DCT. The watermark is added to the image by manipulating the frequency coefficients in the DCT. With complicated algorithms, the area of the picture with the least “noise,” or amount of different shades of color, is chosen to hide the information relating to the watermark. This form of watermarking changes the brightness of the image slightly, and is not visible to the human eye, only the software used to create the watermark. 4. DIGITAL WATERMARKING METHODS There are a variety of techniques to watermark a digital image. The following are descriptions of two of the less complex and more easily understood models. The first algorithm is called least significant bit manipulation. This may seem like a simple algorithm, but it has been proven to be effective at hiding the watermark successfully. The first step in this is to choose how many bits you want to hide the secret image in [6]. This beginning step can make or break how well the algorithm works, the more bits used to hide the secret image, the more the image original image will break down. However, the more bits used to hide the secret image the more likely you will get all of the information out. After you have chosen the number of bits to represent the secret image, you must create a new image by combining the most significant bits of each picture. The most significant bits of the original will stay as the most significant bits in the combined version. The most significant bits of the secret image will become the least significant bits in the combined version. Not Watermarked Figure 6: DCT algorithm used Watermarked When research was started on digital watermarking , it was found that it is a very broad field. A large majority of the information that was found were research papers that were abstract and helped find specific areas of interest that helped write the paper. First of all, it was necessary to delve into the background of watermarking in general to see how it was used in the past, and how it continues to be used today. After more research, visibility aspects of digital watermarking and what the stipulations are that make powerful watermarking algorithms needed to be discovered. Visibility leads to robustness and a great deal more understanding of the subject as a whole. After finding more information on how a watermarking algorithm is made and used, research papers were analyzed in depth and with more force. However, they proved to be very technical, with few or no examples. Digital watermarking will only grow as companies keep finding new and better ways to protect information. . 5. PROPOSED WORK Research has found many watermarking algorithms that are visible, invisible, robust, and fragile. However there is always room for improvement. Future work in the field of digital watermarking should try to encompass multiple layers of watermarks. This is because a visible watermark is unreliable and is not able to fully protect against attacks from users on the internet. The problem with an invisible watermark is that the person viewing the webpage where media is displayed may or may not know if it is copyright protected or not. It is up to the owner of the web page to provide valid documentation that states the image is under copyright protection. Therefore having two layers of watermark, visible and invisible, will allow for a direct knowledge to the user visiting the website that the image is under copyright, and should not be tampered with. A way in which this could be done is with some watermarking algorithms that are already in existence. More research will be needed to determine which aspects of certain watermarking techniques will be retained, but after a short time coding will be able to be undertaken. Some specific watermarking techniques that have been simple to understand may be a stepping stone in working towards an interface that will accomplish the goal of applying both a visible and invisible watermark on a digital image. One major problem that may hinder progress on this project is what mark to put on the image first. If a plan is devised to put an invisible fragile watermark on the image, the visible watermark must be put in place first. This is because any alteration to the digital image after the invisible watermark is put in place destroys it. As a first scheme for a method to implement the said proposal, a visible watermark would be placed on the image. This would be done with a program like gimp or photoshop that has the ability to filter opacity levels in order to not disrupt the image much as a whole. Some testing has already been done in this area as seen in figure 6. In order to decrease time spent on formatting in photoshop a user interface that loops through pictures would be more suitable. Figure 6 Personal Library After a plan is found that cycles through photos in order to put a visible watermark on the image, the invisible watermark needs to be put in place. In realistic terms the watermark would be a fragile watermark, due to its fast execution time. As found in previous research a hash function will have to be devised that manipulates the LSB of each pixel in the image [9]. The general form of the function to do this might look similar to other research in the fragile watermarking field. L(i,j) = fR(R(i,j))fG(G(i,j))fB(B(i,j)) for each pixel (i,j) f:COLOUR(i,j): {7 MSBs} {0,1} (hash function with a key) f:fcolour(binary) : LSB substituted colour value Colour = {R, G, B} i,j : integer variables. Figure 7: Fragile Watermarking Algorithm [9] By implementing the watermark in Figure 7 on a digital image that a visible watermark has been placed on, more security is placed on the image. 5.1 QUALIFICATIONS This project would be able to be done due to experience using programming languages such as python. During time spent being a teacher’s assistant in the class Computing with Images I have become accustomed with programming concepts involved in pixel manipulation. Further experimentation would be needed in order to handle the project at hand, but with my experience teaching the subject, I believe I have a firm grasp on the task at hand. Along with programming experience from classes such as Analysis of Algorithms and Computer Systems the proposed work would be able to be accomplished. 5.2 TIMELINE During the first two weeks of the project further research for invisible watermarking algorithms that may be implementable would be done. This may add work onto the end of the project, but more understanding of past problems and solutions is needed before coding is actually done. The next stage in the proposed work would be to devise and implement an algorithm that cycles through digital images in order to place on a visible watermark. The implementation of this algorithm would be the next two weeks of the timeline. After the first month is complete, hopefully all research for the invisible watermark and coding for the visible watermark will be complete. Once this is done the next two weeks will be allocated to implementing the invisible watermarking algorithm that has been found to be the most suited for the project at hand. If all goes according to the plan stated above, the project will be complete in six weeks. 6. CONCLUSION When research was started on digital watermarking , it was found that it is a very broad field. A large majority of the information that was found were research papers that were abstract and helped find specific areas of interest that helped write the paper. First of all, it was necessary to delve into the background of watermarking in general to see how it was used in the past, and how it continues to be used today. After more research, visibility aspects of digital watermarking and what the stipulations are that make powerful watermarking algorithms needed to be discovered. Visibility leads to robustness and a great deal more understanding of the subject as a whole. After finding more information on how a watermarking algorithm is made and used, research papers were analyzed in depth and with more force. However, they proved to be very technical, with few or no examples. Digital watermarking will only grow as companies keep finding new and better ways to protect information. REFERENCES [1] Mohanty, Saraju P. Watermarking of Digital Images. Rep. INDIAN INSTITUE OF SCIENCE, 1999. [2] Akers, Bob. "History of Watermarks." Kunstpedia. 2009. Web. 5 Nov. 2009. <http://www.kunstpedia.com/articles/454/1/History-ofWatermarks/Page1.html>. [3] Delp, Edward. "Multimedia Security Research at Purdue University." College of Engineering, Purdue University. Digimarc. Web. 5 Nov. 2009. <http://cobweb.ecn.purdue.edu/~ace/water2/digwmk.html #image_adaptive>. [4] Lin, Eugene T., and Edward J. Delp. A Review of Fragile Image Watermarks. Rep. West Lafayette: Purdue University. [5] Kundur, and Hatzinakos. Digital Watermarking for Telltale Tamper-Proofing and Authentication. Tech. Toronto: Natural Sciences and Engineering Research Coucil and Canada, 1999. [6] Cummins, Jonathan, Patrick Diskin, Samuel Lau, and Robert Parlett. "Steganography And Digital Watermarking." School of Computer Science - School of Computer Science. 2004. Web. 12 Nov. 2009. <http://www.cs.bham.ac.uk/~mdr/teaching/modules03/sec urity/students/SS5/Steganography.htm>. [7] Voyatzis, G., and I. Pitas. The Use of Watermarks in the Protection of Digital Multimedia Products. Rep. Thessaloniki: University of Thessaloniki, 1999. [8] Mintzer, Fred, Jeffrey Lotspiech, and Norishige Morimoto. "Safeguarding Digital Library Contents and Users: Digital Watermarking." D-Lib Magazine. IBM Research Division, Dec. 1997. Web. 12 Nov. 2009. <http://www.dlib.org/dlib/december97/ibm/12lotspiech.ht ml>. [9] Lim, Yusuk, Changsheng Xu, and David D. Feng. "Web based Image Authentication Using Invisible Fragile Watermark." Web based Image Authentication Using Invisible Fragile Watermark. Department of Electronic & Information Engineering, Hong Kong Polytechnic University, 2002. Web. 1 Dec. 2009. <http://delivery.acm.org/10.1145/860000/858382/ p31lim.pdf?key1=858382&key2=9015240621&coll=GUI DE&dl=GUIDE&CFID=65981482&CFTOKEN=638715 07>.