There have been some achievements in steganalysis recently. Many people have been making strides in steganalysis. They have approached steganalysis from different angles; from information theory to complexity theory. Anderson gave a possibility that there is a provable secure steganographic system, but there had not been complexity theoretical approaches for years. In 2002, Katzenbeisser and Petitcolas defined the conditional security of steganography and gave a possibility for a practical, provable secure steganography for the first time, and Hopper et al. introduced a provable secure steganographic algorithm in the sense of complexity theory. Chang et al. also tried to define the complexity theoretical security and showed a practical, provable secure algorithm. Chang et al. presented chose-input attack model in a view of pseudoprocessingness for a steganographic system. In this paper, we try to improve this notion in detail. So we define chosen-cover attack model and chosen-message attack model. Moreover, we present the relation between them.