Study for detection of PVD-based Steganography

• Main Authors: Yi, I-Hao, 李翊豪
• Other Authors: Liu, Chiang-Lung
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/93173688047002324348

碩士 === 國防大學理工學院 === 電子工程碩士班 === 100 === Because of 911 tragedy and the requirement of protection of intellectual property rights, many researchers have been devoted to the study of information hiding and steganalysis. The Pixel Value Differencing steganographic method (PVD) adjusts the values of pixel pairs to embed the secret message. Compared with the least singnificant bit replacement method (LSB), the PVD can effectively increase the hiding capacity and avoid the detection of x2 and RS attacks. Afterward, various PVD based information hiding techniques have been proposed to improve the hiding capacity or visual quality of the stego images, including the MPVD and HPVD methos. Although PVD based information hiding techniques can increase the hiding capacity and avoid the detection of x2 and RS attacks, the stego images created by these techniques have abnormal characteristics in the histograms of pixel value differences. Based on these characteristics, many researchers proposed various specific steganalitical mathods to attack PVD based methods. However, there have been no universal techniques that can be used to effectively detect different PVD based methods simultaneously. The thesis aims to propose a universal steganalitical method which can be used to detect and distinguish the stego images created by PVD, MPVD and HPVD. The proposed detection method first proposes three features based on the histogram of pixel value differences. The proposed features are then fed into a Support Vector Mechine (SVM) to decide whether the test image is the cover image or the stego image created by PVD, MPVD, or HPVD methods. Experimental results show that the average accuracy of detection rates provided by the proposed steganalitical method is above 99%. Experimental results also show that, although the proposed steganalitical method is a universal one, the performance for detection of specific PVD based method provided by the proposed method is superior or close to those provided by other specific methods. Therefore, the proposed steganalitical method is practical for protection of information of national defense.