欢迎来到天天文库
浏览记录
ID:36603086
大小:6.51 MB
页数:119页
时间:2019-05-12
《入侵检测的神经网络方法》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、电子科技大学博士学位论文入侵检测的神经网络方法姓名:刘贵松申请学位级别:博士专业:计算机系统结构指导教师:章毅20071201摘要析静态结构的神经气网络缺陷的基础上,提出了一种简单的自增长神经气算法SGNG,分析了该算法的特点,基于此算法给出了两类划分的类别特征刻画,通过建立“normal”数据的特征模式,将之应用于监督异常检测。为了适应输入空间数据的拓扑结构,结合PCA和神经气思想提出了PCNG聚类方法,并给出了其在线学习算法,研究了基于PCNG的入侵检测性能:(4)研究了实际环境的IDS设计问题。基于本文研究的多种方法,
2、提出了针对外部入侵的多种神经网络融合的入侵检测系统模型MNNIDS以及基于ARP协议的内部入侵检测、阻击和内部访问控制方法。对MNNIDS的各个部分功能以及使用的神经网络方法进行了具体设计,分析了其应用特点;对内部入侵检测和访问控制部分进行了实际的软件设计和实现。关键词:入侵检测,PCA神经网络,自组织神经网络,神经气网络,访问控制IIABSTRACTSince1980s,theresearchonthetheoryandapplicationsofartificialneuralnetworkshasalwaysbeena
3、hottopicinthescienceandengineeringfields.Asaveryimportantresearchbranchofcomputationalintelligence,alotofachieve-mentshavebeenobtainedduringthelasttwodecades.Neuralnetworkshavethespecialstructuresandprinciplesforknowledgerepresentationandinformationprocessing.Allth
4、esemeritsresultinlotsofdistinguisheddevelopmentsinmanyapplicationdomains.WiththedevelopmentoftheIntemet,moreandmoreattentionhasbeendrawntonetworksecuritywhichaLsobecomesaattractivefocusformanyre-searchers.Followingthetendencyoflargescalenetworkandcomplexlyintrusive
5、behaviors,theconventionalnetworksecuritytechnologies,aimingatdefencepur-pose,cannotfulfillthenewrequirementsanymore.Therefore,activeprotectingtechnologiescomeintobeingandcanbeusedtotacklethisproblem,includingthemostimportanttechnologyforintrusiondetection.Theresear
6、chonapplyingneuralnetworktothefieldofintrusiondetectionhasattractedmoreandmoreattentionofworldwideresearchers.Combiningtheadvantagesofneuralnetworkswithpracticalcharacteristicsofintrusiondetection,manynewdetectionapproachescanbeconstructed.Studyingthecombinationcan
7、notonlyextendtheapplicationfieldsofneuralnetworks,butalsoacquiremuchsocialandeconomicbenefits.Themaincontributionsofthedissertationareasfollows:(1)UsingPCAneuralnetworks(PCANN)tostudyintrusiondetection.CombiningthepropertiesofPCAforfeatureextractionanddimensionalit
8、yre-duction,theclassifierdesignmethodsaredescribedindetail,includingtheparam-eterssettingmethodsofPCANN-basedclassifier.Afteranalyzingtheshortage
此文档下载收益归作者所有