高级检索

    高深, 林家骏. 基于净图的多模纹理空域隐写分析[J]. 华东理工大学学报(自然科学版), 2011, (6): 754-758.
    引用本文: 高深, 林家骏. 基于净图的多模纹理空域隐写分析[J]. 华东理工大学学报(自然科学版), 2011, (6): 754-758.
    GAO Shen, LIN Jia-jun. A Multimode Texture Spatial Steganalysis Based on CleanImage[J]. Journal of East China University of Science and Technology, 2011, (6): 754-758.
    Citation: GAO Shen, LIN Jia-jun. A Multimode Texture Spatial Steganalysis Based on CleanImage[J]. Journal of East China University of Science and Technology, 2011, (6): 754-758.

    基于净图的多模纹理空域隐写分析

    A Multimode Texture Spatial Steganalysis Based on CleanImage

    • 摘要: 提出了一种基于净图的多模纹理空域隐写分析方法。首先将图片进行切割并按照梯度能量划分成不同的集合;然后对这些集合中的图片用小波镜像滤波器恢复和进行LSB二次攻击,并提取三维净图特征向量;最后通过Oneclass SVM进行训练,得到多超球体SVM。该方法的优点是训练不需要载密图像,对空域隐写分析有很好的效果,在嵌入率为100%、50%、20%时,平滑组分别有100%、99.73%、66.22%的检出率,粗糙组分别有97.6%、72.6%、29.4%的检出率。

       

      Abstract: A multimode texture spatial domain steganalysis based on cleanimage is proposed. Firstly, the image is divided into different sets according to the gradient energy. And then, the images in these sets are restored by using wavelet QMF filter, and attacked by secondLSB steganography, respectively. Finally, a 3D feature vector is extracted, which is further trained by oneclass SVM such that a hypersphere SVM is obtained. The advantage of this proposed method is that the stego images are not required during the training, and have excellent result in spatial domain. When the embedding rate is 100%, 50%, and 20%, the smooth set has the classification rate of 100%, 99.73%, and 66.22%, and the rough set has the classification rate of 97.6%, 72.6%, and 29.4% , respectively.

       

    /

    返回文章
    返回