摘要
为了提高卷积神经网络(CNN)在图像隐写分析领域的分类效果,构建了一个新的卷积神经网络模型(steganalysis-convolutional neural networks,S-CNN)进行隐写分析。该模型采用两层卷积层和两层全连接层,减少了卷积层的层数;通过在激活函数前增加批量正规化层对模型进行优化,避免了模型在训练过程中陷入过拟合;取消池化层,减少嵌入信息的损失,从而提高模型的分类效果。实验结果表明,相比传统的图像隐写分析方法,该模型减少了隐写分析步骤,并且具有较高的隐写分析准确率。
In order to improve the recognition effect of convolutional neural networks( CNN) in image steganalysis,this paper constructed a new steganalysis-convolutional neural networks model( S-CNN) for steganalysis. The model reduced the number of layers of the convolution layer by using two layers of convolution layer and two layers of the whole connection layer. By adding the batch normalization layer to optimize the model before the activation function,to avoid the model in the training process into the over-fitting. The cancellation of the pool layer reduced the loss of embedded information,thereby improving the classification effect of the mode. The experimental results show that,compare with the traditional steganalysis methods,the proposed model reduces the steganalysis step and has higher steganalysis accuracy.
引文
[1]任洪娥,常春武,张健.改进的双线性插值算法在信息隐藏中的应用[J].计算机应用研究,2010,27(11):4290-4292.(Ren Honge,Chang Chunwu,Zhang Jian. Application of improved bilinear interpolation algorithm in information hiding[J]. Application Research of Computers,2010,27(11):4290-4292.)
[2]陶然,张涛,平西建.基于纹理复杂度和差分的抗盲检测图像隐写算法[J].计算机应用,2011,31(10):2678-2681.(Tao Ran,Zhang Tao,Ping Xijian. Based on texture complexity and difference,an anti blind detection image steganography algorithm[J]. Computer Application,2011,31(10):2678-2681.)
[3] Kodovsky J,Fridrich J,Holub V. Ensemble classifiers for steganalysis of digital media[J]. IEEE Trans on Information Forensics&Security,2012,7(2):432-444.
[4] Luo Weiqi,Huang Fangjun,Huang Jiwu. Edge adaptive image steganography based on LSB matching revisited[J]. IEEE Trans on Information Forensics&Security,2010,5(2):201-214.
[5] Pevny T,Bas P,Fridrich J. Steganalysis by subtractive pixel adjacency matrix[J]. IEEE Trans on Information Forensics&Security,2010,5(2):215-224.
[6] Li Bin,Huang Jiwu. Textural features based universal steganalysis[C]//Security,Forensics,Steganography,and Watermarking of Multimedia Contents X. Security,Forensics,Steganography,and Watermarking of Multimedia Contents X. 2008:681912-681912-12.
[7] Qian Yinlong,Dong Jing,Wang Wei. Deep learning for steganalysis via convolutional neural networks[C]//Proc of International Society for Optical Engineering. 2015:94090J-94090J-10.
[8] Fridrich J,Kodovsky J. Rich models for steganalysis of digital images[J]. IEEE Trans on Information Forensics&Security,2012,7(3):868-882.
[9] Kodovsky J,Fridrich J,Holub V. Ensemble classifiers for steganalysis of digital media[J]. IEEE Trans on Information Forensics&Security,2012,7(2):432-444.
[10]Xu Guanshuo,Wu Hanzhou,Shi Yunqing. Structural design of convolutional neural networks for steganalysis[J]. IEEE Signal Processing Letters,2016,23(5):708-712.
[11] Sainath T N,Mohamed A R,Kingsbury B,et al. Deep convolutional neural networks for LVCSR[C]//Proc of IEEE International Conference on Acoustics,Speech and Signal Processing. Piscataway,NJ:IEEE Press,2013:8614-8618.
[12]Cirstea B I,Likforman-Sulem L. Improving a deep convolutional neural network architecture for character recognition[J]. Electronic Imaging,2016,2016(17):1-7.
[13]Vuckovic F,Lauc G,Aulchenko Y. Normalization and batch correction methods for high-throughput glycomics[C]//Joint Meeting of the Society-For-Glycobiology. 2015:1160-1161.
[14]Tan Shunquan,Li Bin. Stacked convolutional auto-encoders for steganalysis of digital images[C]//Proc of IEEE Signal and Information Processing Association Summit and Conference. Piscataway,NJ:IEEE Press,2014:1-4.
[15]Tang Weixuan,Tan Shunquan,Li Bin,et al. Automatic steganographic distortion learning using a generative adversarial network[J]. IEEE Signal Processing Letters,2017,24(10):1547-1551.