Research And Implementation Of Steganography Analysis Of Digital Image Based On Deep Learning

In recent years,with the rapid development and application of Internet information technology,people have become accustomed to transmitting information by means of the Internet.Therefore,information security issues have received more and more attention.Digital steganography is an effective means of protecting secure communications by embedding information into a carrier to enable secure transmission of information.However,this technology is like a double-edged sword.While protecting our secure communications,it can also be used by illegal elements to use illegal activities.Therefore,it is obviously of great significance to have a steganalysis technique for detecting whether steganography is applied.With the help of high-performance GPU training,convolutional neural network enables many algorithms based on machine learning and depth learning to be applied to various industries.In the field of image,deep learning has become a hot topic in recent studies.Therefore,it has become a trend to apply the success of deep learning in the field of image classification to the analysis of image steganography.Combined with the characteristics of image steganography and the advantages of general deep learning image classification.In combination with the success of deep learning in the field of image steganography analysis in recent years,this paper introduces the research topic of this paper,the research and implementation of digital image steganography analysis based on deep learning,and conducts the following specific research on the characteristics of steganography analysis:(1)the existing main image steganography and image steganography analysis methods and algorithms are summarized,and several typical deep learning models are studied in depth.The idea and method of solving image steganography by deep learning are introduced.Several commonly used steganographic analysis filters are introduced.(2)in view of the particularity of the image steganography analysis task,a special design is made for the universal deep learning network,and the accuracy and speed of the analysis are studied by changing the network structure,the number of layers of the network,the activation function and adding the convolution of the filter layer.Experimental analysis was carried out for different embedding rates.The effectiveness of the deep convolutional neural network in detecting image steganography is verified.(3)in order to reduce the training time,a method using migration learning is proposed,which directly calls the pre-trained VGG16 network structure on ImageNet.The pre-trained parameter weights are used as the initial network parameters of the previous layers,and then the weights of the later layers of the network are directly trained.The experiment shows that the speed of the pre-training method is obviously improved when the accuracy is not seriously lost,compared with the situation that each layer has training parameters.