Abstract: 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.

Key words: steganalysis, image steganography, hiding information, fast Fourier transform(FFT), quantum noise

摘要： 隐写术和隐写分析是同一枚硬币的两个不同面，两者都同样重要。图像隐写被认为是最有前途的安全数据传输方法之一，因为它将数据隐藏在图像文件中。相反，隐写分析试图攻击隐写术，并检索可能是秘密消息的隐藏数据。许多鲁棒和强大的图像隐写方法已经在文献中提出。其中一种方法被称为无嵌入隐写术（SWEM），被认为更安全，因为采用了数据传输的概念，而不是将其嵌入或隐藏在文件中。像任何新技术一样，图像隐写术可能会产生负面影响，并可能被滥用。对于用SWEM方法构造的图像文件，目前还没有能够识别和攻击的隐写分析方法。为了克服上述问题，本文主要贡献是提出了一种基于快速傅里叶变换（FFT）的图像隐写分析新方法。提出的方法首先使用FFT检查可疑图像。然后，通过计算得到的FFT频谱的不同部分的比例，并与阈值进行比较，可以判断可疑图像是否被SWEM构建。该方法克服了现有其他方法无法攻击这种强图像隐写方法的缺点。为了检验本文方法作为一种识别和攻击由SWEM生成的图像文件的方法的泛化性，构建了一个图像数据集，使用SWEM算法构建未使用嵌入术的隐藏秘密信息图像，并将本文方法与其他两种常见的隐写分析方法进行了比较。结果表明：与另两种方法相比，所提出的方法能够对这些图像进行识别，并且具有较高的识别率。

关键词: 隐写分析，图像隐写术，隐藏信息，快速傅里叶变换，量子噪声