Let's continue our discussion of IP protection with digital watermarking. This cartoon shows how watermarking can be used as a deterrent for IP piracy. Alice is IP provider. And Bob gets a copy of the IP from Alice. Seeing the value of the IP, Bob decides to make profit illegally from Alice's IP. He can reverse engineer the IP, and reproduce it or he can also integrate the IP into his own product. Then, Bob can sell the I, reverse engineered IP or his new design to another user without the authorization from Alice. This type of IP infringement can happen very easily because of the IP we use, particularly with the soft IPs. It is also very hard to detect and trace. Even when Bob is caught, Alice may not have sufficient evidence to prove her authorship of the IP. Now suppose Alice embeds her signature as watermark into the, into the IP. After Bob reproduces the IP or integrates the IP into his own design, before he can resell it, he has to think about whether he has removed Alice's signature or watermark completely. If not, he might get caught and face penalty because this time Alice can prove her authorship of the IP from the watermark or from the, her signature. Therefore, we see that watermark can deter attackers from illegally copying or using IPs. Digital watermarking has been widely used for identification, and notation and copyright of multimedia data such as text, image, audio, and the video. Traditional watermarking techniques take advantage of the limitations of human visual and auditory systems and embed the signature to the original data as small er, errors. This actually changes the original multimedia data and cannot be directly applied for the protection of hardware design intellectual properties. This is because the value of these IPs rely on their correct functionalities and their performance. We observe that for a given system specification, there are normally many different ways to implement the system, or develop the IPs. Therefore, if the IP designer can intentionally create certain structures in the implementation of the IP to make it rather unique, then this evidence can be used as watermark to prove IP's authorship. For example, this is the gate-level view of two implementations of a four-bit arithmetic and logical unit. On the left is the original design. The one on the right is upturned after we embed the message UMCP TERPS in the original varial code. We can easily see the different appearance of the two implementations. For example, here, here, and here. But it is hard to tell which one has watermark and which one does not have. The next example is a FPGA implementation of the data encryption standard. One with no watermark, another one with a message, more than 4700 bits embedded. Again, we see difference of the two designs here and here. Now we analyze the constraint based watermarking from the perspective of steganography system. Here is the classical steganography system. Secret data is hidden in the stego data, behind the cover data, using a stego key. The secret can also be extracted from the stego data, if necessary. Here is a similar drawing showing the concept of the constraint-based watermarking system. The original problem for the system specification, specification of the IP, will be used as the cover constraint to hide the watermark and author's signature. To do this, the IP designer will first create a set of constraints. These constraints are selected in such a way that they do not conflict with the cover constraints. This is because conflicting constraints my turn the problem into unsolvable. Then the original constraint and this additional constraint will be combined together to form a stego-problem. The stago-problem, instead of the original problem, will then be solved to obtain a stego-solution. This stego-solution will satisfy both the original and the additional constraints. To claim the authorship, the designer has to demonstrate that the Stegno solution carries information based on his signature here. The Stegno solution necessarily satisfies a set of additional constraints. These constraints may look random. However, the designer can regenerate these constraints using his signature together with his Stego-key. Cryptograph, cry, cryptography functions, such as one-way hash function and stream cipher, can be used to generate this embedded constraints to enhance the credibility of the authorship. Finally, we consider I, IP invention as solving a hard problem. The requirements for the IP are the constraints that any solution is to meet to be a solution for the problem. The value of the IP will be measured by the quality of the solution. We now show the concept of charting solution space behind the constraint-based watermarking approach. This green area shows a solution space to our original problem. The purple are in the middle shows the solution space of better solutions. This point is a randomly found, found the solution to the problem. After we add extra constraints some of the solutions failed to satisfy these additional constraints and they become invalid. Suppose this shaded area is the solution space after watermarking. Now, if we report the same solution it is more convincing to say that we find this solution by solving it, by solving the watermarked problem instead of solving the original problem. To see this, suppose there are, there are n solutions to the original problem and m solutions to the watermarked solutions. When n is greater, when n is greater than m or in many cases when n is much greater, much larger than m, the chance that finding a solution from the m solutions, which is 1 over m, will be much bigger than the chance of finding the solution from n solutions. And this can be used as a very strong proof of the authorship of this reported solution, which is in some sense the authorship of the IP.
