Information hidden
Information hidden model information hidden (INFORTION HIDING) Unlike traditional cryptography technology. The password technology is mainly to study how special information is specially encoded to form an unrecognized password form (ciphertext) to transfer; and information hide mainly study how to secretly hide some confidential information in another disclosed information. The confidential information is then passed through the transfer of public information. For encrypted communication, a possible monitaker or illegal interceptor can be accepted by intercepting ciphertext, or deciphering it, or sends the ciphertext, thereby affecting the security of confidential information; but for information hiding, Possible monitors or illegal interceptors are difficult to determine whether confidential information is present from the public information, and it is difficult to intercept confidential information to ensure the security of confidential information. The wide application of multimedia technology provides a broader field for the development of information hiding technology.
We call the hidden information to secret information (SECRET Message), which can be copyrighted information or secret data, or a serial number; the public information is called a carrier information, such as video, audio clip. This information hidden process is typically controlled by the key ("Embedding Algorithm), which is hidden in public information, while the concealed carrier (the public information of secret information) is passed through the channel (Communication Channel) ) Pass, then detector (Detector) recovers / detects secret information from the hidden carrier.
Information hiding technology is mainly composed of two parts: (1) information embedding algorithm, which utilizes a key to hide the secret information. (2) Concealed Information Detection / Extraction Algorithm (detector), which detects / restores secret information from the hidden vector using a key. Under the premise of the unknown key, the third party is difficult to get or delete from the hidden carrier, and even discover secret information.
Information hidden feature information is hidden than traditional encryption because its purpose is not to restrict normal data access, but to ensure that hidden data is not violated and discovered. Therefore, information hiding techniques must consider the threats caused by normal information operations, that is, the confidential data is immunized to normal data operation technology. The key to this immunity is to make the hidden information part is not easily damaged by normal data operations (such as usual signal transform operations or data compression). According to the purpose and technical requirements of information hidden, this technology has the following characteristics:
Robustness refers to the ability to hide information loss due to some changes in the image file. The so-called "modified" here includes channel noise, filter operation, heavy sample, lossless coding compression, D / A or A / D conversion, or the like during transmission. UndeteTECTABILITY means that the hidden vector has a consistent characteristic of the original carrier. If there is a consistent statistical noise distribution, the illegal interceptors cannot determine whether there is hidden information. Transparency (Invisibility) utilizes human visual systems or human auditory system properties, through a series of hidden processing, making the target data no significant degradation, but hidden data can't see or hear it. Security (Security) refers to the hidden algorithm has strong anti-attack capabilities, that is, it must be able to withstand some degree of person-assault, but make hidden information will not be damaged. Self-recovery, due to some operation or transformation, it may cause a large destruction of the original map, and if only the segment data left, the hidden signal can still be restored, and the recovery process does not require a host signal, which is the so-called self Restore. Information hidden is an emerging cross-discipline, has a broad application prospect in computer, communication, and confidentiality. Digital Watermarking Techniques As its important application in multimedia, it has received more and more attention.
digital water mark
Background With the development of digital technology and the Internet, various forms of multimedia digital works (images, videos, audio, etc.) have been published in the form of a network, and their copyright protection has become an urgent need to solve problems. Since Digital Watermark is a valid approach to copyright protection, it has become a hot spot in the field of multimedia information security research, and is also an important branch in the field of information hiding technology. This technique confirms the ownership of the data by embedding a secret information-watermark in raw data. This embedded watermark can be a paragraph, logo, serial number, etc., and this watermark is usually invisible or inevitable, it is closely combined with raw data (such as image, audio, video data) and hides it. And you can save some operations that do not destroy the source data use value or commercial value can be saved. In addition to the general characteristics of information hiding technologies, digital watermarking technologies also has its own intrinsic features and research methods. In the digital watermark system, the loss of hidden information means the loss of copyright information, thereby losing the function of copyright protection, that is, this system is failed. It can be seen that the digital watermark technology must have strong robustness, security, and transparency. Typical Digital Watermark System Model Figure 1 is a watermark signal embedded model, which is functional to complete the addition of the watermark signal to the original data; FIG. 2 is a watermark signal detection model to determine whether a specified watermark signal is contained in a certain data.
Digital watermarking main application
Copyright Protection, that is, the owner of the digital works can generate a watermark and embed its original data, and then publicly publish his watermark version. When the work is pirated or a copyright dispute, the owner can use the method of obtaining a watermark signal from the pirated work or watermarking article using FIG. 3 or FIG. 4, thereby protecting the owner's interests. Add fingerprints to avoid unauthorized copy making and distribution, producers can use different users' ID or serial number as different watermarks (fingerprints) embedded in the legal copy of the works. Once an unauthorized copy is found, it can determine its source according to the referred to this copy. The title is embedded in the form of the work in the form of the work in the form of the work in the form of the article (e.g., shooting time and location, such as a photo), such as a photo, etc.), and this implicit annotation does not require additional bandwidth and is not easily lost. Tamper Tips When the digital work is used for court, medicine, news, and business, it is often necessary to determine whether their content is modified, forged or specially handled. To achieve this, the original image can usually be divided into multiple separate blocks, and then add each block to different watermarks. At the same time, the integrity of the work can be determined by detecting the watermark signal in each data block. Unlike other watermarks, such watermarks must be fragile, and when the watermark signal is detected, it does not require raw data. A typical example of using controlling this application is a DVD copy system, i.e., adding watermark information to DVD data so that the DVD player can determine its legitimacy and copyability by detecting watermark information in the DVD data. To protect manufacturers' business benefits.
Typical Digital Water Printing Algorithm has made great progress in recent years, and some typical algorithms have been analyzed below, except for image data (some algorithms are also suitable for video and audio) data).
Airspace algorithm. The typical watermark algorithm in this class algorithm is to embed information onto the least important pixel bit (LSB: Least Significant Bits) in the randomly selected image point, which ensures that the embedded watermark is invisible. However, since the image is not important, the robby of the algorithm is used, and the watermark information is easily destroyed, the image is quantized, and the geometric deformation is destroyed. Another common method is to use pixel statistical features to embed the brightness value of the information in the pixel. The PatchWork algorithm method is randomly select N to pixel points (AI, BI), and then add the brightness value of each AI point to 1, and the brightness value of each BI point is reduced, so the average brightness of the entire image remains unchanged. Adjust the parameters properly, the PatchWork method has certain resistance to JPEG compression, FIR filtering, and image cropping, but the amount of information embedded in this method is limited. In order to embed more watermark information, the image can be subjected to block, and then each image block is embedded. Transform domain algorithm. In this class algorithm, most of the watermarking algorithms use the Spread Spectrum Communication technology. The algorithm implementation process is: first calculate the discrete cosine transform (DCT) of the image, and then add watermark to the maximum first K coefficient of the magnitude of the DCT domain (excluding DC components), usually the low frequency component of the image. If the first k maximum component of the DCT coefficient is represented as d = {di}, i = 1, ..., k, watermark is a random real number of random real numbers that obey Gaussian distribution w = {wi}, i = 1, ..., K, then water printing The embedded algorithm is di = di (1 AWI), where constant A is a scale factor, controlling the strength of the watermark added. Then use the new coefficient to do the reaction to obtain the watermark image I. The decoding function calculates the discrete cosine shift of the original image I and the watermark image i *, and extracts the embedded watermark W *, and then performs the relevant test to determine whether the presence or absence of the watermark. This method can still extract a trusted watermark copy even if the watermark image has been more significantly deformed by some universal geometric deformation and signal processing operation. A simple improvement is not to embed the watermark to the low frequency component of the DCT domain, but embedded to the intermediate frequency component to adjust the contradiction between the recruitment of watermarking and invisibility. In addition, the spatial domain data of the digital image can be converted into the corresponding frequency domain coefficient by the discrete Fourier transform (DFT) or discrete wavelet transform (DWT); secondly, according to the type of information to be hidden, it is appropriately encoded. Or deformation; again, based on the size of the hidden information and its corresponding security objectives, select some type of frequency domain coefficient sequence (such as high frequency or intermediate or low frequency); again, determine a certain rule or algorithm, to be hidden The corresponding data of the information modifies the previously selected frequency domain coefficient sequence; finally, the frequency domain coefficient of the digital image is converted into spatial domain data with respective reverse transforms. The hidden and extraction information of this class is complex, and the amount of hidden information cannot be large, but strong anti-attack ability is very suitable for digital watermarking technology for copyright protection of digital works. The compressed domain algorithm based on JPEG, the MPEG standard compressed domain digital watermark system not only saves a lot of complete decoding and re-encoding process, but also has a lot of practical value in digital TV broadcast and VOD (Video On Demand). Accordingly, watermark detection and extraction can also be carried out directly in the compressed domain data. A digital watermark scheme for MPEG-2 compressed video data stream is described below.
Although the MPEG-2 data flow syntax allows the user data to be added to the data stream, this scheme is not suitable for digital watermarking techniques, as user data can be removed from the data stream while the video data stream is encoded at the MPEG-2 Increase user data will increase bit rate, so that it is not suitable for the application of fixed bandwidth, so how is the key to add watermark signals to the data signal, i.e., add to the data stream representing the video frame. For the input MPEG-2 data flow, it can be divided into data head information, motion vector (for motion compensation) and DCT encoded signal block 3, only the last part of the MPEG-2 data stream is changed in the scheme, The principle is to first decode and inverse each of the HUFFMAN codes in the DCT encoded data block to obtain a DCT coefficient of the current data block; secondly, the conversion coefficient of the corresponding watermark signal is added to it, thereby obtaining The DCT coefficient of the watermark is superimposed, and then the quantization and Huffman encoding. Finally, the number N1 of the new HUFFMAN codeword is compared to the original watermark coefficient of code N0, only when N1 is not more than N0, can transfer watermark Code word, otherwise transmit the original code word, which guarantees that there is no video data stream bit rate. This method has a problem to consider, that is, the introduction of the watermark signal is an error signal that causes degradation, and the coding scheme based on the motion compensation will cause an error to solve this problem, the algorithm has taken drift compensation The solution offsets visual deformation caused by the introduction of the watermark signal. NEC algorithm This algorithm is proposed by a NEC laboratory Cox et al. Poses an important role in digital water printing algorithms. It is a method of producing a pseudo-random sequence by a key as a seed, which has a Gaussian N (0, 1) Distribution, the key is generally composed of the author's identification code and the hash value of the image, followed by the image to DCT transform, and finally use a pseudo-random Gaussian sequence to modulate (superimpose) This image except DC (DC) 1000 largest DCT coefficients. This algorithm has strong robustness, safety, transparency, and the like. Due to the special key, IBM attack can be prevented, and the algorithm also proposes an important principle of enhancing the watermark roll resistance and anti-attack algorithm, i.e., the watermark signal should be embedded in the source data. The watermark signal is composed of independent distribution random real numbers, and the real number sequence should have the characteristics of Gaussian distribution N (0, 1). Physiological model algorithm's physiological models include human vision systems HVS (Humanvisualsystem) and human auditory system HAS. This model is not only utilized by multimedia data compression systems, but also available for digital watermarking systems. The basic thinking of the visual model is to determine the maximum intensity of the digital watermark signals that can be tolerated from the various parts of the image from the JUST NOTICEABLERENCE described by the visual model, thereby avoiding the damage of visual quality. That is, the visual model is used to determine the modulation mask associated with the image, and then use it to insert the watermark. This method has good transparency and strongness.
Conclusion Information Hidden and Digital Watermarking Technology is a frontier research area rise in the international academic community in recent years. It has a close relationship with information security, information hiding, data encryption. Especially in today's rapid development of network technology and application, the research of watermarking technology is more practical. In the future, the research of watermarking technology will still focus on recreational, authenticity differentiation, copyright certification, network fast automatic verification, and audio and video watermarking, and will closely combine with data encryption technology, especially in recregnance and provenibility Research. The drilling of watermarks reflects the survivability of watermarking in digital files. Although the current most algorithms have certain recreational, if various image attacks are applied, then these algorithms will fail. How to find a more mupping watermark algorithm is still an urgent need to solve problems. In addition, the current watermark algorithm is more incompletely provided by providing reliable copyright certificates, so finding a digital watermarking algorithm that provides full copyright protection is also an important research direction. There are currently a lot of watermarking methods (see Resources), the author will discuss another article for attack methods and corresponding solutions. Reference
Related Readings Beijing University of Posts and Telecommunications Information Security Center Yang Yizhen Professor "Information Hidden Technology - Interpute and Digital Watermark"
Related sites
Ezstego, Stego Online, Stego
