Histogram Modification of Pixel Differences (HMPD) is an important reversible data hiding algorithm. HMPD distributes the value of adjacent Pixel Differences, and uses the highest values on Histogram to hide the information. Overall, HMPD's data hiding shows good effect, but as for the commonly seen steganalysis tool, Subtractive Pixel Adjacency Matrix (SPAM), HMPD is unable to avoid the detection of SPAM effectively. We use RSP, BOWS-2. RSP85, etc. image database to cany out this experiment. In this paper, in the SPAM training process, we use different embedded sheets to detect whether HMPD can effectively avoid the detection of SPAM. The results show that no matter the number of sheets embedded in the training process, SPAM can still effectively detect whether images use the HMPD method to embed information. We can see from the above that the number of embedded images in the SPAM training process do not influence its ability to distinguish. In the future we can work toward to embed a small amount of information to test whether SPAM can effectively detect it.

In the age of the information, while people conveniently obtaining and exchanging the information, how to prove the ownership of information is become an important issue. Watermark can be some words, signs and order numbers etc. and in the digital watermark, the loss of the embedded information is the same as the loss of the rights. Above all, digital watermark must have more powerful robustness and no perceptibility. Thus, we based on Least Significant Bit (LSB) to propose the algorithm of high robustness and high transparency. This algorithm conducted repeatedly embedding on the high bit of the watermark image and used the algorithm of the LSB data hiding at the same time. In the experiment, we used salt and pepper noise to test the robustness that we proposed, and used the steganalysis technology of Steganalysis by Subtractive Pixel Adjacency Matrix (SPAM), which is proposed by Fridrich et al. in 2010 to detect the no perceptibility. The method we proposed can be seen to effectively discriminate the watermark after being attacked in this experiment and it owns no perceptibility.

A large amount of information is transmitted on the Internet, some don't want to be seen by others, how is data-hiding in an image as not to be seen by the eye or pass through stcganalyzcr tools, for example, Subtractivc Pixel Adjacency Matrix (SPAM) detect possible? This has become an important problem. Chao et al. proposed A Novel Image Data Hiding Scheme with Diamond Encoding way of data hiding, shortened as DE, DE has high capacity and whether information is hidden is not easily seen by the eye. As stcganalysis's development matures, DE is unable to pass the stcganalysis technology, SPAM. As a result, DE has lost its data hiding use, so in this research, we discuss how to improve DE's camouflage ability. The way DE embeds is from left to right, from up to down, putting two elements in one group and changing these elements to embed information, this is also easier to embed information into continuous areas, and thus, easily detected by SPAM. As a result, this research proposed a way to disperse information randomly and hiding the information in a dispersed way in the image to strengthen the ability of camouflage. The results of this experiment show that by using what we proposed, it can effectively lower the possibility of being detected.

During the same time as data mining, it may lead to leakage of personal privacy issues; k-anonymization algorithm is usually used to protect personal privacy, but they are very complicated algorithms, when k-anonymity is a NP-hard problem, k is equal or greater than 3. Traditional clustering algorithm is used to improve the computational efficiency of k-anonymization, but it takes a higher time cost. In order to improve the abovementioned problems, we introduce a fast and stable clustering algorithm, to improve the k-anonymous operation and receive better results. The method of time complexity is only O ( n * q ) , better than the One-Pass K-means Algorithm of O ( n 2 / k ) , where n is the sampling number of data, q is the number of quasi-identifier and k is the number of anonymities. Proven via the experiment, and with k = 50 as the example, this method execution time only takes 16 seconds, the One-pass K-means Algorithm execution time is 120 seconds. This study provides a fast and stable result of the k-anonymity algorithm.

Pixel Pair Modification (PPM) based magic squares is proposed for data hiding. The aim is to improve the drawbacks of payload versus PSNR of LSB. The traditional PPM-related methods did not consider the combination of square contents. If we can give the best combination of the magic square, making the cover image pixel value for the least amount of change. So the same amount of possession, but also embed the highest performance. As the size N × N permutation and combination of the magic square that is (N × N)! Species almost not can be effectively solved. This paper size for the N × N optimal combination of the magic square problem of the development of an exhaustive algorithm. To 3 × 3 magic square the size of the experiment, using the image signal to noise ratio Peak signal of noise ration (PSNR) as a image quality camouflage to determine performance benchmark. Method in the construction of a data matrix and pixel square difference between mobile data matrix, to quickly exhaustive combination of all of its solutions. The sequence if all of the confidential information embedded into the cover image, only extract pixel pair from the cover image. Search for magic square, revise pixel pair and restore to the cover image. This process as the standard embed course. In the study of results to showing, this paper's method is explanation the first combination solution and compare the standard embed course. It can reduce the solution time of about 68.7%.About 3.37s.And the other of the combination solution. Each combination of solutions to solve only the standard embed the process 0.076%.About 0.0082s. Therefore, this paper method can effectively solve the time-consuming by brute-force method. Quickly to get Optimal solution of the magic square, it can be used to enhance performance of pixel pair modification (PPM) based magic squares is proposed for data hiding.

An Efficient Prediction-and-Shifting Embedding Technique for High Quality Reversible Data Hiding(EPSET) is based on the subtraction between the pixel itself and the average value of the pixel which is around the pixel point. Is a high-capacity reversible data hiding, EPSET, the resulting difference between the distribution of hidden data in the histogram, this approach has a high capacity and cannot image the naked eye whether the embedded data. Steganalysis technique is to detect whether the image which the data embedded in a technology, Among them, Fridrich et al. in 2010 proposed Steganalysis by Subtractive Pixel Adjacency Matrix (SPAM) steganalysis technique is more effective. We use RSP this image database and in different capacities to test EPSET whether can escape SPAM detection, EPSET can be drawn from the experimental data at low volume you still cannot effectively escape the SPAM detection. The main reason is EPSET Embedded data on smooth region, SPAM makes it easier to detect, Future, we will test EPSET embed more complex areas can escape detection SPAM.