Article Preview
TopIntroduction
Network communication and information technology have developed rapidly in the era of increasing net-workization. Cloud computing has provided enough space for individuals and enterprised to store multimedia data(Khan et al., 2018). Users can use the cloud to store and share data. There are two ways to prevent image information from leaking: encryption and information hiding (Guo et al., 2011). Encryption technology of image is to ensure the security of images, which uses digital image matrix features to change pixels according to the transformation rules of images in space or transform domain to achieve encrypted values (Liu et al., 2013; Wang et al., 2005; Samidha & Agrawal, 2013; Hemalatha et al., 2013). However, it will cause image distortion and make the image into a form of noise or texture, which may cause suspicion of the attacker and increase the possibility of information leakage, loss and tampering (Dang & Chau, 2000). The significant information was embed into the carrier by modifying the carrier data (e.g. digital image, video, audio, etc.) to realize the hiding of the important information. The procedure of information hiding avoids the attention of attackers (Sakkara & Somashekar, 2012; Zhou & Chen, 2006; Zhang et al., 2003). Moreover, digital image contains a number of information, and it is the most widely used as an ideal information hiding carrier and it is the most widely used as an ideal information hiding carrier (Liang&Wang, 2007; Qian&Xu, 2018; Qian & Zhang, 2016; Wang & Zhang, 2002; Ren&Sui, 2002). Technology of the classical image information hiding constitutes of spatial domain information hiding and transforms domain information hiding. The methods of the spatial domain information hiding include Least Significant Bit (LSB)(Chan & Cheng, 2004), Adaptive Least Significant Bit (Yang&Weng, 2008), Pixels Value Differencing (PVD)(Wu&Tsai, 2003), S-UNIWARD(Holub&Fridrich, 2014), and WOW (Holub & Fridrich, 2012) etc. The methods of the transform domain include Discrete Fourier Transform (Chaumont & Puech, 2006), Discrete Cosine Transform (Cox&Kilian, 1996), and Discrete Wavelet Transform (Lin&Horng, 2008) information hiding and so on. All of these methods embed secret information according to certain rules through modifying the vector, and consequentially leave modified marks on the vector. Therefore, the information hiding methods mentioned above are difficult to resist detection of various steganalysis methods. The principles of typical information hiding have been shown in Figure 1.
Figure 1. Typical model of information hiding
Objective to effectively resist the test of steganalysis methods, Zhou et al. used the new conception of coverless information hiding, and compared with the classic information hiding method, the coverless information hiding doesn’t need to encode the significant information into the carrier (Zhou&Cao, 2016). In (Zhou&Cao, 2016), the bag-of-words model is used to extract the visual keywords of the image. To indicate information that needs to be hidden, this realizes the hiding of the text information in the image. The method wouldn’t the modify the carrier. However, a large number of codebook needs to be built, which has large storage overhead and small hiding capacity. In (Zhang&Qian, 2016), the structural information hiding is proposed, but the object library that synthesizes the dense carrier needs to be segmented from a large number of normal image libraries, so the method is inefficient. This paper proposed coverless information hiding based on WGAN-GP model. The scheme doesn’t change cover image and can be effectively prevented from being detected by the steganography analysis tool. The work mainly includes the following points.