项目名称: 图像隐写的定量与定位分析若干问题研究
项目编号: No.61302159
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 无线电电子学、电信技术
项目作者: 杨春芳
作者单位: 中国人民解放军信息工程大学
项目金额: 28万元
中文摘要: 隐写定量与定位分析是隐写取证研究的重要组成部分,能够为秘密信息的提取和还原提供帮助。然而,现有定量分析方法对多位平面替换隐写的嵌入率估计精度有待于提高,且难以适用于自适应隐写;定位分析方法的前提条件较为严格。本课题针对图像多位平面替换隐写和干扰量化类自适应隐写的定量分析、空域图像和JPEG图像隐写的定位分析等,拟利用多元高斯模型和支持向量回归,对多位平面替换隐写的多种嵌入率估计方法进行融合;剔除不可更改的统计样本,运用支持向量回归估计干扰量化类隐写的系数更改比率;选取疑似被更改像素子集,依据子集中的像素更改比率判定选取的像素是否被空域图像隐写更改;利用相邻分块系数、小波去噪后系数与载体系数间的相关性,对载体系数进行估计,根据估计的载体系数与待检测图像系数的差异定位JPEG图像隐写的被更改系数。通过这些研究,期望在相关问题上有所突破,为秘密信息的提取和还原提供理论和技术上的支持。
中文关键词: 隐写;隐写定量分析;隐写定位分析;隐写信息提取;自适应隐写
英文摘要: Quantitative and locating steganalysis are the important components of steganography forensics, can provide help to the estraction and recovery of secret message. However, the existing quantitative steganalysis methods would estimate the embedding ratio of multiple bit planes replacement steganography with higher accuracy, and are not applicable to the adaptive steganography, the existing locating steganalysis methods work under strict precondition. This project focuses on the quantitative steganalysis of image multiple bit planes replacement steganography and perturbed-quantization-like adaptive steganography, and the locating steganalysis of spatial image steganography and JPEG image steganography. The following researches are planned to be carried out: adopt the multivariate Gaussian model and support vector regression to fuse different quantitative steganalysis methods of multiple bit planes replacement steganography; eliminate the unchangeable statistical sample, then estimate the coefficient modification ratio of perturbed-quantization-like adaptive steganography by support vector regression; select possible modified pixel subset, then judge whether the selected pixels have been modified based on the pixel modification ratio in the selected pixel subset; estimate the cover coefficients by utilizing the cor
英文关键词: steganography;quantitative steganalysis;locating steganalysis;hidden message extraction;adaptive steganography