指纹向量在垃圾邮件检测中的应用  

SPAM FILTERING METHOD BASED ON FINGERPRINT

作　　者：刘卫红[1] 方卫东[1] 张凌[1] 董守斌[1] 

机构地区：[1]华南理工大学计算机学院,广东广州510641

出　　处：《陕西科技大学学报（自然科学版）》2009年第5期84-88,共5页Journal of Shaanxi University of Science & Technology

基　　金：国家下一代互联网示范工程基金资助项目(CNGI-04-12-2A)

摘　　要：大量重复发送的邮件是垃圾邮件的主要生成方式,为检测这种类型的垃圾邮件,作者提出了一种基于指纹向量的自适应垃圾邮件过滤方法.该方法中,每封邮件通过q元取样后映射成指纹向量,如果两个邮件的指纹向量的距离较小,则认为其属于同一个类别.该方法具有计算速度快、占用计算资源小的优点,与传统的贝叶斯算法相比具有较优的性能,在TREC2007数据集的实验结果验证了所提出方法的有效性.Unsolicited bulk mail is the spam′s main form.In order to detect spam of this form,an adaptive spam filtering mechanism based on message fingerprinting is presented in this paper.In our mechanism,each message is represented by a fingerprint vector after q grams sampling procedure,and two messages with a short distance in their fingerprint vectors are considered as variants of each other.It is more fast and takes less computer resource.Comparing with Bogo filtering method it is with better performance also,experiments on TREC2007 dataset demonstrate the effectiveness of the proposed method.

关 键 词：指纹向量 q元 垃圾邮件 

分 类 号：TP393.098[自动化与计算机技术—计算机应用技术]