基于免疫Agent的垃圾邮件过滤模型  

A SPAM FILTERING MODEL BASED ON IMMUNE-AGENT

作　　者：蒋亚平[1] 田月霞[1] 梅骁 

机构地区：[1]郑州轻工业学院计算机与通信工程学院,河南郑州450001

出　　处：《计算机应用与软件》2016年第3期294-298,313,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61272038);河南省科技厅科技攻关项目(0624220084)

摘　　要：针对传统的垃圾邮件过滤方法不能有效识别未知特征及变异特征、终端服务器负载较大和接收邮件时延较长等问题,借鉴生物免疫学原理和多Agent技术,设计一种基于免疫多Agent垃圾邮件过滤模型SF-MA。该模型通过对SMTP协议改进,可快速地判断垃圾邮件的产生,并记忆特征信息;设计抗原提呈算法,扩大自体库的规模;将疫苗概念引入模型,保留优良基因,实现各个Agent的信息交互,增强了整个模型"记忆"机制,有效地提取垃圾邮件的信息和变异特征。利用邮件样本集对该模型进行训练和测试,仿真结果表明,该模型与其他模型相比具有更好的性能,有效地提高了垃圾邮件模型的正确率等特性,降低了虚报率。For the problems of traditional spam filtering methods such as cannot effectively identifying the unknown and variation features,heavier load in terminal server and longer delays in receiving mails,etc.,by making use of biological immunology principle and multi-Agent technology,we designed an immune multi-Agent-based spam filtering model SF-MA. The model is able to quickly judge spam generation and to remember the feature information by improving the SMTP protocol; we also designed an antigen presentation algorithm,and expanded the scale of self library; we introduced the vaccine concept into model to keep good genes and to realise the interaction of each Agent ＇s information,these enhanced the ＂memory＂mechanism of entire model,and effectively extracted the information and variation features of the spam. Using the mail sample set to train and test the model,simulation results showed that the proposed model had better performance than other models,and effectively improved the characteristics of accuracy rate of spam model,as well as reduced false alarm rate.

关 键 词：人工免疫 SMTP协议 垃圾邮件 抗原提呈 疫苗 多AGENT 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]