基于Kernel Fisher Discriminant的JPEG文件隐形信息检测算法  

A JPEG Steganalysis Algorithm based on Kernel Fisher Discriminant

作　　者：潘晓声[1] 黄继风[1] 

机构地区：[1]上海师范大学,上海200234

出　　处：《微计算机应用》2005年第5期567-570,共4页Microcomputer Applications

摘　　要：隐秘信息分析的目的是正确的分辨出文件中是否带有隐藏的信息。自然图像和隐秘图像的DCT系数经过差分之后,它们的直方图会呈现不同的规律,以此为特征向量,来检测JPEG图像是否带有隐藏信息。并通过对低隐藏信息量的JPEG图像进一步分析,发现采用线性核函数的核Fisher判别有着较好的检测性能。实验结果表明,本文的算法有效。The purpose of image steganalysis is to find out whether the hidden messages exist in images. After differencing DCT coefficients of natural--images and stego--images, their histograms present different laws. This paper regards this as the feature vector to determine whether a hiding message is in the image. Then we further analyse JPEG image which hides the low amount of information, and find that using Kernel fisher discriminant with linear kernel function has good capability of classifying. The results of the experiments show our approach is effective.

关 键 词：数字信息隐写术 隐秘分析术 自然图像 隐秘图像 FISHER线性判别 核FISHER判别 FISHER判别 JPEG文件 检测算法 信息分析 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]