The behavior and nature of attacks and threats to computer network systems have been evolving rapidly with the advances in computer security technology. At the same time however, computer criminals and other malicious...The behavior and nature of attacks and threats to computer network systems have been evolving rapidly with the advances in computer security technology. At the same time however, computer criminals and other malicious elements find ways and methods to thwart such protective measures and find techniques of penetrating such secure systems. Therefore adaptability, or the ability to learn and react to a consistently changing threat environment, is a key requirement for modern intrusion detection systems. In this paper we try to develop a novel metric to assess the performance of such intrusion detection systems under the influence of attacks. We propose a new metric called feedback reliability ratio for an intrusion detection system. We further try to modify and use the already available statistical Canberra distance metric and apply it to intrusion detection to quantify the dissimilarity between malicious elements and normal nodes in a network.展开更多
P2P网络中的节点很可能从另外的节点那里收到质量很差的服务和信息,名誉评价是解决该问题的常见方法.基于评分反馈的P2P名誉计算机制存在下述缺点:无法区分恶意评价和诚实节点给出错误评价间的差别;需要对评分可信度进行二次评价,使名...P2P网络中的节点很可能从另外的节点那里收到质量很差的服务和信息,名誉评价是解决该问题的常见方法.基于评分反馈的P2P名誉计算机制存在下述缺点:无法区分恶意评价和诚实节点给出错误评价间的差别;需要对评分可信度进行二次评价,使名誉计算速度减慢;用数字来表示节点名誉的方式不够自然.实际上,名誉评价的用途是确定节点可信度的相对顺序.因此,提出了一种基于排名反馈的P2P名誉评价机制RbRf(reputation based ranking feedback).针对RbRf和其上的恶意攻击进行了数学建模和理论分析,结果表明,RbRf中非恶意错误的影响随排名反馈的数量指数而衰减;一般恶意攻击对RbRf的影响随排名反馈数量的多项式而减小;对于有意设计的共谋攻击,由于必须给RbRf引入正确信息而导致了恶意攻击被有效中和.因此,RbRf不仅由于不再反馈打分信息而不存在评分反馈引起的名誉评价问题(如不需要对反馈信息的可信度进行二次评价),而且具有更好的抵抗恶意攻击的能力.仿真实验验证了理论分析的结果.展开更多
文摘The behavior and nature of attacks and threats to computer network systems have been evolving rapidly with the advances in computer security technology. At the same time however, computer criminals and other malicious elements find ways and methods to thwart such protective measures and find techniques of penetrating such secure systems. Therefore adaptability, or the ability to learn and react to a consistently changing threat environment, is a key requirement for modern intrusion detection systems. In this paper we try to develop a novel metric to assess the performance of such intrusion detection systems under the influence of attacks. We propose a new metric called feedback reliability ratio for an intrusion detection system. We further try to modify and use the already available statistical Canberra distance metric and apply it to intrusion detection to quantify the dissimilarity between malicious elements and normal nodes in a network.
基金Supported by the National Basic Research Program of China (No. 2007CB311201), the National Natural Science Foundation of China (No.60833008 No.60803149), and the Foundation of Guangxi Key Laboratory of Information and Communication (No.20902).
文摘P2P网络中的节点很可能从另外的节点那里收到质量很差的服务和信息,名誉评价是解决该问题的常见方法.基于评分反馈的P2P名誉计算机制存在下述缺点:无法区分恶意评价和诚实节点给出错误评价间的差别;需要对评分可信度进行二次评价,使名誉计算速度减慢;用数字来表示节点名誉的方式不够自然.实际上,名誉评价的用途是确定节点可信度的相对顺序.因此,提出了一种基于排名反馈的P2P名誉评价机制RbRf(reputation based ranking feedback).针对RbRf和其上的恶意攻击进行了数学建模和理论分析,结果表明,RbRf中非恶意错误的影响随排名反馈的数量指数而衰减;一般恶意攻击对RbRf的影响随排名反馈数量的多项式而减小;对于有意设计的共谋攻击,由于必须给RbRf引入正确信息而导致了恶意攻击被有效中和.因此,RbRf不仅由于不再反馈打分信息而不存在评分反馈引起的名誉评价问题(如不需要对反馈信息的可信度进行二次评价),而且具有更好的抵抗恶意攻击的能力.仿真实验验证了理论分析的结果.