136 The Steganalysis of Histogram Modification of Pixel Differences Based on Spam Data Mining

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.