一种基于Q学习的LDoS攻击实时防御机制及其CPN实现.pdf

计算机研究与发展 ISSN1000—12391CN 11—1777／TP of Journal Researchand 48(3)：432—439，2011 Computer Development 一种基于Q学习的LDoS攻击实时防御机制及其CPN实现 刘 陶h4 何炎祥h2 熊 琦1’3 1(武汉大学计算机学院武汉430072) 2(软件工程国家重点实验室(武汉大学)武汉430079) 3(中国信息安全评测中心北京100085) 4(幸福人寿保险公司信息技术部北京100027) (1iutao6l@126．corn) A BasedReal—Time LDoSAttackand Mechanism Q—Learning Mitigating against Its andSimulationwithCPN Modeling Liu Ta01”，He Qil·3 Yanxian91～．andXiong School，Wuhan 430072) 1(Computer University，Wuhan 2(State 430072) KeyLaboratoryofSoftwareEngineering(WuhanUniversity)，Wuhan Evaluation 100085) 3(ChinaInformation Security Center。Beijing Insurance 100027) ‘(HappyLife Co．，Ltd，Beijing AbstraetDifferentfromtraditionaIDoS andlOW- attacks，lOW—rateDoS(LDoS)iS stealthy，periodic in hard in rateattackvolume，andis tobedetectedanddefended thesefeatures very time．Regarding ofLDoSattack，we areal—time mechanismbasedon forLDoSattack． present mitigating Q—learning the control asthe of module Takingadaptationsystem targetprotecting，aQ—learningimplemented withBP-neutral takescharacteristicextracted fromnetwork network，which parameters periodically asits usedtOmakechoiceofthebestdefense theniS input，iS