A Watermarking System Based On Digital Signature And Digital Timestamp

Nowadays,the rapid development of computer and networkhave greatly changed people's life. Many inventions andproductions are stored or transported by digital ways. Andmultimedia information exchange based on computer and networkalso offer the convenience for the usage of digital works. However,digital works prevailing on the net are easily captured,duplicatedand juggled , which seriously impinges the copyright of collectivityand personage. In order to solve this problem, as a powerfulmeasure to protect the ownership, digital watermarking isconceived. So-called digital watermarking is a technique of embeddingsome valuable information about copyright in digital works. Thus,it can protect the ownership. In a total digital watermarking system,the loss of watermarks means the loss of ownerships. Obviously,system like this fails to protect the copyright. So it is necessary thatthe digital watermarking must possess some characteristics as thefollowings: 1, Invisibility (or clarity): that is, the original work mustn'tbe reduced its quality after a series of hiding processings, while thehiding data can't be seen or listened by people. 2, Robustness: It is the most important characteristic of all.It refers to that the whole system can resist some signal processingmeans and even intended attacks. What's more, the watermark canstill be drawn correctly.3, Certainty: The watermark associating ownership shouldbe exclusively identified.By taking account of the domain in which the watermark isembedded, watermarking schemes can be classified into two broadcategories: spatial-domain and frequency-domain techniques. Thesimplest watermarking techniques embed the watermark directlyinto the spatial domain by modulating the least-significant bit planeof the original image. Modulating the least-significant bit, thehuman visual organ is not sensitive to the difference because thechange in intensity of certain pixels is small. The advantages ofspatial-domain techniques are high perceptual transparency andefficiency. The disadvantage is fragile to image processing andgeometric translations. The frequency-domain watermarkingtechnique embeds the watermark by modulating the magnitude ofcoefficients in a transform domain, such as DCT, DWT, DFT. Inconsideration of this robustness, we adopt the frequency domainscheme in concert with public key cryptography under a reasonablecomputing computing complexity.Wavelet analysis is an active researching field in mathematicsand engineering. Comparing to traditional Fourier analysis,wavelet analysis has very good time-frequency character in signalprocessing. Moreover, it has the fast decomposition andreconstruction algorithm of Mallat. With the theory of waveletsteadily perfected, its applications become widely into many fieldssuch as signal and image processing,pattern recognition,computervision and graphics. We briefly review the DWT model: The imageis first decomposed into four subbands denoting LL1 , LH 1, HL1 andHH 1. LH 1, HL1 and HH 1 contain the finest scale detailedwavelet coefficients, i.e., the higher frequency detailed information.LL1 , the coarse overall shape, is the low frequency componentcontaining most of the energy in the image. The wavelet transformis then applied to obtain the next coarser scale by furtherdecomposing LL1 in to LL2 , LH 2, HL2 and HH 2.If the process isrepeated T times, we can obtain the subband LLT . In the humanvisual system, people are more sensitive to low frequencycomponents than high frequency components. Under reasonableattacks, the low frequency components can survive.In past years, the proposed watermarking techniques areprivate watermarking schemes that use the same secret key forembedding and extraction. The main advantage is that theembedded watermark is difficult to remove without thecorresponding secret key. Unfortunately the secret key is revealedwhen proving the existence of the watermark in the test image.Once the watermark is verified, the secret key must be discardedbecause the released information is enough to deduce the secret key.To eliminate this weakness, the owner must embed differentwatermarks into the same original image using different secretkeys. It is difficult to manage numerous watermarks and thecorresponding secret keys. Collusion/averaging attacks are possiblewith this type of system. However, these drawbacks can be avoidedby introducing public-key cryptography into the digitalwatermarking scheme. In a dispute involving a copyright or patent:if two persons claim ownership, how does the notary justify whopossesses the exact ownership? Of course, the person that producesthe earliest copy of the disputed work wins the case. Justice can beachieved using two well-defined security services: time-stampingand digital signature. Time-stamping is a technique used to verifywhether digital data was created or signed at a certain time. On theother hand, a digital signature is a piece of information based onboth the signed data and the signer's private key. In such a way,anybody can verify the signature with access to public key of thesigner. In a word, the digital signature provides the solution foridentifying the rightful author and digital time-stamp for when thedata was created. Hence, to protect something that you want topatent, a combination of the digital timestamp and digital signatureprovide an effective solution. However, most digital data such asimages, audio and video have the characteristics of easyduplication and allowing reasonable distortion. A private canutilize these characteristics to imperceptibly modify original digitaldata without perception through the human senses alone and claimownership of that digital data. The reasonable distortions make nodifference to the human eye, but fail the security serviceverification tools. Because these services are sensitive to the bitstream rather than the content, we need to combine the digitalsignature and digital time-stamping techniques into digitalwatermarking system.In this paper, we proposed two algorithms aimed at binarywatermark images and Gray-level logo images respectively. First,we apply DWT to the host image, get the T scale coarse subbandLLT (C for short). Then, according to some rules, we get R, whohas the same size of C, from C, we do some operation between Cand W to get K. Finally we sign K and W with digital signature anddigital time-stamping technique.We define the comparability NC as below to judge the hostwatermark and the drawn one:For binary watermark images:NC (W , W ) = n /NFor Gray-level watermark images:2 2( , )( )( )( ) ( )NC W WW W W WW W W W=? ?? ?∑∑ ∑In addition, using the PSNR to estimate the distortion degreeof the host image:210psnr10 log ( 255)= ? MSEhere, 2,1 [ ( , ) ( , )]MSE = 512 × 512 ∑i j I i j ?I i jExperimental results show our algorithms are invisible androbust against many signal processings such as cropping,compression, filtering and noise adding. So they are effective andpractical.Of course, the algorithms remain to be improved for they stillhave some flaws, for example, their robustness to rotation is weak.That will be the further research work in my future.