基于截图内容的图片垃圾邮件过滤系统  

A spam image filtering system based on user-specified image content

作　　者：陈俊伟[1] 张丽春[1] 吕岳[1] 

机构地区：[1]华东师范大学计算机科学与技术系,上海200241

出　　处：《智能系统学报》2008年第5期416-422,共7页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金资助项目(60475006);教育部新世纪优秀人才支持计划资助项目(NCET-05-0430)

摘　　要：垃圾邮件制造者常常将文字嵌入到图像中,产生了大量的图片垃圾邮件.为解决这一问题,提出并实现了一个基于截图内容的图片垃圾邮件过滤方案.首先由用户从垃圾邮件中截取某一子域图片,每一截图对应一类垃圾图片,所有的截图构成一个自定义的垃圾图片"黑名单".其次对读入的每一封图片邮件,其内置图片与"黑名单"中的图片进行图像匹配.最后若存在匹配项,则判定该邮件含有用户已指定的垃圾图片信息.将此图片垃圾邮件过滤方案应用于一个小型的邮件收发系统,使用3 534幅垃圾邮件图片进行实验,结果证明了该垃圾邮件过滤方案有效.Spammers often embed text in images in innocuous seeming emails, resulting in large numbers of spam images which are difficult to filter out. To solve this problem, a spare image filtering system was developed that applies user-specified identification of spare image content. First, users were asked to identify spain images from emails to generate a class of spam images. These images were then added to a blacklist. Next, images in incoming e-mails were matched with images stored in the image blacklist. Finally, if a matching image was found, the e-mail was judged as spare because it contained images identified as spam by users. Testing with an e-mail server containing 3534 sample images proved the proposed method effectively intercepts spare e-mails.

关 键 词：截图内容 图片垃圾邮件过滤 图像匹配 

分 类 号：TP393.098[自动化与计算机技术—计算机应用技术]