摘要
最低有效位(least significant bit,LSB)替换,有顺序和随机嵌入2种方式,现有的检测方法大多只针对其中一种方式。为了实现对顺序和随机嵌入的同时检测,采用数理统计的方法对灰度图像嵌入信息前后小波高频系数的变化进行分析,建立了相应的线性回归模型,并采用基于内容的图像检索技术和机器学习的方法对模型进行了优化,以此来估计嵌入信息的长度。与RS(regular singular)和Pairs典型检测算法相比,该方法不仅能检测明文和密文嵌入,而且对嵌入信息长度估计的准确性有所提高,实验结果表明绝对误差的均值不大于2.033%。
Secret messages can be embedded sequentially or randomly in images using least significant bit (LSB) replacement. Based on the difference of the high-frequency wavelet coefficients between cover-images and stego-images, this paper use linear regression model to detect messages hidden in gray scale images by LSB replacement. Combining context-based image retrieval technology and machine learning, the accuracy of the length estimate can be improved. Experiment results demonstrate the mean of absolute error is not greater than 2.03%. Compared to RS and Pairs steganalysis methods, our algorithm can reliably detect both encrypted and unencrypted messages, while the two typical methods can only detect the encrypted ones.
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2007年第4期595-598,共4页
Journal of Tsinghua University(Science and Technology)
基金
国家自然科学基金资助项目(90304014)
关键词
LSB替换
隐秘分析
线性回归模型
小波变换
LSB(least significant bit) replacement
steganalysis
linear regression model
wavelet transform