一个基于TSVM的GIF图像通用隐写检测方法  

An Universal Steganalysis Method for GIF Image Based on TSVM

作　　者：钟尚平[1] 林静[1] 

机构地区：[1]福州大学数学与计算机科学学院,福州350002

出　　处：《北京交通大学学报》2009年第2期122-126,共5页JOURNAL OF BEIJING JIAOTONG UNIVERSITY

基　　金：国家"863计划"项目资助(2005AA142110);福建省青年人才基金项目资助(2006F3076)

摘　　要：直推式支持向量机(TSVM)是一种从标记样本出发,对特定的未标记样本进行识别和分类的技术.本文分析了将TSVM应用到图像通用隐写检测中的可行性,提出并实现了基于TSVM的GIF图像通用隐写检测方法.实验结果表明,针对不同的GIF图像隐写工具,本文方法在较少标记样本条件下引入大量未标记样本,得到接近监督学习的通用检测效果.从而提高了GIF图像通用隐写检测的实用性.Transductive support vector machine （TSVM） is a technology which starts with labeled samples and aims at recognizing and classifying the given unlabeled samples. In this paper, the feasibility of applying TSVM to universal image steganalysis was analyzed, and a universal steganalysis method for Gif image based on TSVM were proposed and implemented. The experimental results shown that, in the condition of few labeled samples and massive unlabeled samples, this method can achieves the approximate detecting effect of old method based on traditional inductive support vector machines. Thus, the model of universal steganalysis for Gif image based on TSVM is more practical.

关 键 词：信息安全 通用隐写检测 直推式支持向量机 GIF图像 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]