图像型垃圾邮件过滤技术综述  

Survey on image-based spam filtering

作　　者：万明成[1] 耿技[1] 程红蓉[1] 陈佳[1] 

机构地区：[1]电子科技大学计算机科学与工程学院,成都610054

出　　处：《计算机应用研究》2008年第9期2579-2582,共4页Application Research of Computers

基　　金：国家"863"计划资助项目(2006AA01Z411)

摘　　要：从基于图像特征的图像型垃圾邮件的检测难点入手,总结了目前用于识别垃圾邮件的图像特征,将其归类为文件属性、图像属性等八类特征。对已经用于图像型垃圾邮件分类的五种分类算法,包括支持向量机、决策树法、最大熵模型、D-S证据理论、贝叶斯算法进行了理论分析与效果比较。最后对图像型垃圾邮件过滤技术的研究方向进行了展望。This paper analyzed the difficulties of detecting image-based spain with the features of images in detail. The features of spare images were divided into eight categories such as file attribute features, image metadata and so on. Then, discussed and compared five classification algorithms which have been used in image-based spam filtering were outlined, including support vector machines, decision tree method, maximum entropy model, the DS evidence theory, Bayesian algorithm and the effect of these algorithms. Finally,gave some future directions of research on the techniques of image-based spam filtering.

关 键 词：图像型垃圾邮件 人为干扰 垃圾邮件图像 特征分析 分类算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]