论文部分内容阅读
网络生存能力分析主要研究在攻击等意外事件中网络实体正常运行的能力。目前,为搞清某种攻击对网络的影响,通常需要在数千种配置情况下运行数千种场景来测试Ad hoc网络的生存能力。然而,此方法的效果并不能令人信服:配置方式是人为选择或随机的,测试结论不能推广到其它场景。针对这些问题,提出了一种提高生存能力分析效率的新方法,该方法使用机器学习和以攻击者为中心的网络表示法。收集了一个数据集并用其构造了一个分类器,该分类器准确(正确率大于97%)预测了网络欺骗和数据转发攻击造成的数据流损失。
Network Survivability Analysis focuses on the ability of network entities to function properly during unexpected events such as attacks. Currently, in order to understand the impact of an attack on the network, thousands of scenarios typically need to be run in thousands of configurations to test the viability of Ad hoc networks. However, the effect of this approach is not convincing: configuration options are artificially chosen or randomized and test conclusions can not be generalized to other scenarios. In response to these problems, a new method of improving the analysis of viability is proposed, which uses machine learning and attackers-centric network representation. A data set was collected and used to construct a classifier that accurately (with a correctness greater than 97%) predicted the loss of data flow from spoofing and data-forwarding attacks.