Abstract
Image steganalysis is a technique for detecting the presence of hidden information in images, which has profound significance for maintaining cyberspace security. In recent years, various deep steganalysis networks have been proposed in academia, and have achieved good detection performance. Although convolutional neural networks (CNNs) can effectively extract the features describing the image content, the difficulty lies in extracting the subtle features that describe the existence of hidden information. Considering this concern, this paper introduces separable convolution and adversarial mechanism, and proposes a new network structure that effectively solves the problem. The separable convolution maximizes the residual information by utilizing its channel correlation. The adversarial mechanism makes the generator extract more content features to mislead the discriminator, thus separating more steganographic features. We conducted experiments on BOSSBase1.01 and BOWS2 to detect various adaptive steganography algorithms. The experimental results demonstrate that our method extracts the steganographic features effectively. The separable convolution increases the signal-to-noise ratio, maximizes the channel correlation of residuals, and improves efficiency. The adversarial mechanism can separate more steganographic features, effectively improving the performance. Compared with the traditional steganalysis methods based on deep learning, our method shows obvious improvements in both detection performance and training efficiency.
Highlights
Image steganography is a technology that hides secret information in images
We introduced an adversarial mechanism into the network structure to suppress image content information and highlight steganographic information as much as possible
To investigate whether the introduced separable convolution and adversarial training can retain less information about the image content in the extracted features, we removed the separable convolution module and the gradient reversal layer (GRL) from the network structure in order to verify the performance of the network separately
Summary
Image steganography is a technology that hides secret information in images. Due to its simplicity, variability, and difficulty of detection and extraction [1,2], it can be used by illegal organizations to engage in activities that will endanger both national and public security. Ture learning and channel feature learning, thereby maximizing theSeparable channel convolution correlation of Introduction of separable convolution module: splits normal the noise residuals to improveinto thepointwise signal-to-noise ratio in order to detect the subtle differconvolution convolution and depthwise convolution. The image via contains content information reflectlearning and channel feature learning, thereby maximizing the channel correlation of the ing the visual perception of the image, along with steganographic information reflecting noise residuals to improve the signal-to-noise ratio in order to detect the subtle differences the embedding of steganographic messages. We performance use the idea ofof transfer network, withoutfor thereference, interference content adversarial information. By introducing the above two modules, the proposed improves and highlight steganographic information asnetwork much as significantly possible; doing so can better extract the accuracy of steganalysis. Learning for reference, and introduce adversarial training [22] to suppress content inforIntroduction of adversarial training: The image contains content information reflecting mation and highlight steganographic information as much as possible; doing so can better the visual perception of the image, along with steganographic information reflecting the extract steganographic embedding features and improve the learning embedding of steganographic messages.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
