摘要： 针对隐写分析中的难点——空域LSB匹配隐写进行检测和分析，描述LSB匹配加性隐写的特点，将匹配隐写建模为图像受到一定强度的脉冲噪声干扰，采用小波变换对退化的图像进行恢复作为载体的估计。对检测图像和恢复图像提取多个直方图特征比值作为特征向量，利用支持向量机对500幅高质量未压缩的自然图像组成的载体、载密图像库进行检测，结果证明该算法在低嵌入率下可获得较好的检测效果。

关键词: 隐写分析, LSB匹配, 图像恢复

Abstract: The problem of LSB matching steganalysis in grayscale images is considered and a new method for detection is provided. The matching embedding as a kind of image degradation by certain additive pulse noise is modeled, and the restored image by wavelet denoising technology as the estimation of the cover image is used. The ratios of 1D grayscale histogram features of test images and the estimated images as detectors are computed, and the cover and stego images with support vector machine are classified. Experimental results prove that this method is more reliable than the previous related methods.

Key words: steganalysis, LSB matching, image restoration

中图分类号: 

• TP309