基于深度学习的软件定义网络应用策略冲突检测方法

发布时间：2018-09-04 18:24

【摘要】：在基于Open Flow的软件定义网络(SDN)中,应用被部署时,相应的流表策略将被下发到Open Flow交换机中,不同应用的流表项之间如果产生冲突,将会影响交换机的实际转发行为,进而扰乱特定应用的正确部署以及SDN的安全。随着SDN规模的扩大以及需要部署应用的数量的剧增,交换机中的流表数量呈现爆炸式增长。此时若采用传统的流表冲突检测算法,交换机将会耗费大量的系统计算时间。结合深度学习,首次提出了一种适合SDN中超大规模应用部署的智能流表冲突检测方法。实验结果表明,第一级深度学习模型的AUC达到97.04%,第二级模型的AUC达到99.97%,同时冲突检测时间与流表规模呈现线性增长关系。
[Abstract]:In the software defined network (SDN) based on Open Flow, when the application is deployed, the corresponding stream table policy will be sent down to the Open Flow switch. If there is a conflict between the different application stream table items, it will affect the actual forwarding behavior of the switch. This disrupts the proper deployment of specific applications and the security of SDN. With the expansion of SDN scale and the number of applications needed to be deployed, the number of stream tables in the switch increases explosively. If the traditional stream table collision detection algorithm is adopted, the switch will consume a lot of system computing time. Combined with deep learning, a new intelligent stream table collision detection method is proposed for the first time, which is suitable for the deployment of large scale applications in SDN. The experimental results show that the AUC of the first depth learning model is 97.04, and the AUC of the second level model is 99.97. At the same time, the conflict detection time is linearly increased with the scale of the stream table.
【作者单位】：浙江工商大学信息与电子工程学院;美国佛罗里达大学大规模智能系统实验室;
【基金】：国家高技术研究发展计划(“863”计划)基金资助项目(No.2015AA011901) 国家自然科学基金资助项目(No.61402408,No.61379120) 浙江省自然科学基金资助项目(No.LY18F010006) 浙江省重点研发计划基金资助项目(No.2017C03058)~~
【分类号】：TP18;TP393.0

【相似文献】