基于CAPTCHA和Winnow算法的垃圾短信过滤研究  

Research on SMS spam filtering based on CAPTCHA and Winnow

作　　者：王世卿[1] 张燕丽[1] 

机构地区：[1]郑州大学信息工程学院,河南郑州450002

出　　处：《计算机工程与设计》2011年第1期313-315,357,共4页Computer Engineering and Design

基　　金：国家"十一五"科技支撑计划基金项目(2006BAF01A18)

摘　　要：为识别并过滤掉日益增多的垃圾短信,提出了基于全自动人机识别系统(CAPTCHA)和Winnow算法的过滤方法。在CAPTCHA方法中,根据用户能否正确辨认图片,人类和计算机能被辨别,该方法能有效地过滤计算机发送的组垃圾短信。改进的Winnow过滤器可以直接处理原始文本,节省了中文分词时间,而且利用了复合分类思想,提高了分类精度。实验结果表明,CAPTCHA和改进的Winnow算法相结合能较准确地过滤掉垃圾短信。To identify and filter out spam messages whose number is growing,a new filter method based on CAPTCHA and Winnow is proposed.In the CAPTCHA method,humans and computers can be identified according to if the user can identify pictures correctly, this system can filter group-sending-Spam which are from computers effectively.Improved Winnow filter can deal with original text directly, which saved the time of Chinese word segmentation,what＇s more,this paper made use of multiple classifier and classification accuracy is improved.Experiments show that this method can filter out spam messages accurately.

关 键 词：全自动人机识别系统 WINNOW算法 垃圾短信 中文分词 复合分类 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]