Novel Steganalysis Method for Stego-Images Directly Constructed from Color Images Based on Their Quantum Noise

Steganography and steganalysis are two different sides of the same coin. Both are just as important as the other. Image steganography is considered one of the most promising secure data transmission methods because it hides the data in an image file. In contrast, steganalysis tries to attack steganography and retrieve the hidden data that can be a secret message. Many robust and powerful image steganography methods have been presented in the literature. One of these methods is referred to as steganography without embedding (SWEM). It is considered more secure because it adopts the concept of data transmission without embedding or concealing it in the file. Like any new technology, image steganography may have a negative impact and can be misused. There is no steganalysis method for distinguishing and attacking the image file that has been constructed by the SWEM method. To overcome the aforementioned issues, the main contribution of this paper is to propose a new image steganalysis method based on the fast Fourier transform (FFT). The proposed method starts by examining a suspected image using FFT. Then, by calculating the ratio of different parts of the resulting FFT spectrum and comparing the result with a threshold value, we can decide whether the suspected image has been constructed by the SWEM or not. The proposed method overcomes the other existing methods that fail to attack this strong image steganography method. To check the generalizability of the proposed method as a method for distinguishing and attacking the image file that is created by SWEM, we constructed a data set of images, used the SWEM algorithm to construct images with concealed secret messages without embedding, and applied the proposed method and other two common existing steganalysis methods. The results show that the proposed method is able to distinguish these images with a high recognition rate compared to the results of the two methods.