Ad Hoc网络中基于重复博弈的攻击预测模型  

Attack Prediction Model Based on Repeated Game in Ad Hoc Network

作　　者：严辉[1] 沈士根[2] 曹奇英[1] 

机构地区：[1]东华大学计算机科学与技术学院,上海201620 [2]东华大学信息科学与技术学院,上海201620

出　　处：《计算机工程》2012年第7期110-112,115,共4页Computer Engineering

基　　金：国家自然科学基金资助项目(61003278);国家科技支撑计划基金资助项目(2008BAH37B05092)

摘　　要：大多数入侵检测系统预测攻击行为能力较弱,不能实时主动地响应攻击。针对该问题,提出一种基于重复博弈的攻击预测模型。通过建立入侵者和入侵检测系统之间的阶段博弈模型,给出阶段博弈的纳什均衡,并求出重复博弈情况下的子博弈精炼纳什均衡。使用QRE模型预测攻击者和防御者在博弈阶段1～t中选择各种策略的概率,利用Gambit分析得出预测结果。基于GloMoSim的仿真结果表明,相比纳什均衡模型,QRE均衡模型的预测效果更好。Most Intrusion Detection System（IDS） are weak in real-time response attacker＇s action,so it can not predict attacker＇s action well.Aiming at the problem,this paper puts forward an attack prediction model between attacker and IDS based on repeated game.It establishes the stage game model with the two players and gives the Nash equilibrium,gets the unique sub-game perfect Nash equilibrium when the game plays infinity times and chalks up the probability of every strategy of two players in stage 1~ stage t according to the Quantal Response Equilibrium（QRE） model with the Gambit simulation.The model is simulated in Ad Hoc network by using the GloMoSim,whose results show that compared with Nash equilibrium model,QRE equilibrium model gains better prediction effect.

关 键 词：阶段博弈 重复博弈 子博弈精炼纳什均衡 QRE模型 攻击预测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]