基于LSBMR图像边缘自适应隐写的检测  

Detection for Edge Adaptive Image Steganography Based on LSBMR

作　　者：张艳[1] 叶学义[1] 汪云路[1] 鲁国鹏[1] 

机构地区：[1]杭州电子科技大学模式识别与信息安全实验室,杭州310018

出　　处：《计算机工程》2013年第6期205-209,共5页Computer Engineering

基　　金：国家自然科学基金资助项目(60802047;61001216);浙江省自然科学基金资助项目(R1090138);浙江省科技计划基金资助重点项目(2008C21092)

摘　　要：在图像边缘自适应LSB匹配改进隐写算法中,秘密信息嵌入位置的选择仅由某个方向上像素对的差值决定,未考虑该像素与其邻域内其他像素的差值变化的特点。针对该问题,对隐写前后图像的八方向差分直方图进行分析,提出一种基于LSB匹配改进算法(LSBMR)边缘自适应隐写检测的算法。该算法计算图像的八方向绝对差分直方图,提取直方图中隐写前后变化较为明显的频数用以构建特征向量,并使用支持向量机完成检测。对较低嵌入率下(≤0.5 bpp)的EALSBMR隐写结果进行检测,结果表明该算法的平均检测率均高于现有典型的隐写分析算法。Focusing on that the choice of embedding positions within a cover image during edge adaptive image steganography based on LSB Matching Revisited（EALSBMR） depends on the difference between two consecutive pixels on certain direction, without taking into account the difference change between the pixels and its adjacent area, a steganalysis algorithm for EALSBMR is proposed based on comparing the eight-direction difference histogram of cover and stego image. It calculates the eight-direction absolute difference histogram. Then histogram bins changed obviously are built for features. At last, Support Vector Machine（SVM） as classifier output the result of steganalysis. Experimental data for lower embedding rates（≤ 0.5 bpp） illustrate that the proposed method is superior to currently typical steganalysis methods.

关 键 词：隐写分析 边缘自适应隐写 LSB匹配改进算法 八方向绝对差分直方图 支持向量机 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]