基于图像攻击的隐藏信息盲检测技术  

An Attacked Image Based Hidden Messages Blind Detect Technique

作　　者：毛家发[1,2] 林家骏[2] 戴蒙[3] 

机构地区：[1]上饶师范学院数学与计算机系,江西上饶334001 [2]华东理工大学信息与控制学院,上海200237 [3]上海应用技术学院计算机科学与工程系,上海200233

出　　处：《计算机学报》2009年第2期318-327,共10页Chinese Journal of Computers

基　　金：江西省教育厅科技基金项目(Gjj08462);上海市教育发展基金会晨光计划(2008CGB21)资助~~

摘　　要：隐写分析技术是网络信息安全的重要组成部分.它的最根本问题是判断数字载体是否携带秘密信息.文中提出了基于攻击的隐藏信息盲检测方式,提取空域图像、DCT域图像一个二维特征向量,通过非线性神经网络分类器来判别图像是否含有秘密信息,并建立了7000多幅图像库(包括掩密图像与干净图像)进行了可行性、多样性仿真实验,取得了以下效果:(1)平均检测率90.085%(阳性检测率与阴性检测率的总平均);(2)不受限于隐写方案;(3)能够检测低嵌入率的掩密图像;(4)能够检测出经平滑、锐化、缩小、剪切和再压缩处理后的干净图像.Steganalysis technique is an important part of the security of internet information. It＇ essential problem is to estimate if the digital carrier takes the secret messages . This paper brings forward the blind detect technique which is based on the hidden messages for aggressive. A 2D feature vector in the spatial and DCT domains is extracted to blindly determine the existence of hidden messages in an image. For evaluation, a database composed of over 7000 cover and stego images（generated by using ten different embedding schemes） was established. Based on the data- base, simulation experiments were conducted to prove the feasibility and diversity of the proposed system. The proposed system consists of following： （1） The average detection rate is about 90. 085%, the total average for electropositive detection rate and negative detection rate; （2） It is not limited to the detection of a particular steganographic scheme; （3） It has the capability to detect the stego images with an lower embedding rate; （4） The test of plain images which is smoothing, sharpening, image size zoomed out, cutting and recompressing can be inspected.

关 键 词：隐写术 隐写分析 数字水印 数据隐藏 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]