一种检测多类自适应隐写的盲检测方法

自适应隐写算法将秘密消息嵌入到图像纹理复杂的边缘区域,在图像平坦区域不嵌或少嵌.由于图像的复杂纹理能够很好地掩护秘密信息.因此在此区域隐写就更难被检测.针对此问题,用不同模板对图像滤波得到图像噪声残留,提取噪声残留的共生矩阵,并联合由图像水平、垂直及对角线3种不同方向的差分直方图中提取的特征作为分类特征.通过对3种典型的自适应隐写算法的测试,表明该算法能够取得较好的盲检测效果.

Blind Steganalysis of Multi Adaptive Image Steganography

Adaptive steganography makes embedding changes in those regions of the cover image that is hard to model, which makes the detection more difficult. The image noise residual is computed by convolving the image with different kernels to solve the problem. Steganalytic features consist of the fourth order co occurrence matrix calculated from the image noise residual and the features extracted from the pixel value difference histogram with different directions of the image. The proposed steganalytic algorithm is tested on three different typical adaptive steganography methods. Extensive experimental results show that the proposed steganalytic algorithm has superior performance compared with the previous ones.