MNOS：拟态网络操作系统设计与实现

Design and Implementation of Mimic Network Operating System

控制层的漏洞利用攻击,如恶意APP、流表篡改等是软件定义网络(software defined networking,SDN)面临的主要威胁之一,而传统基于漏洞修复技术的防御策略无法应对未知漏洞或后门.提出一种基于拟态防御思想的网络操作系统安全架构——拟态网络操作系统(mimic network operating system,MNOS)——保障SDN控制层安全.该架构采用异构冗余的网络操作系统(network operating system,NOS),并在传统的SDN数据层和控制层间增设了拟态层,实现动态调度功能.首先拟态层动态选取若干NOS作为激活态并行提供服务,然后根据各NOS的处理结果决定最终的有效响应返回底层交换机.实验评估表明:在增加有限的时延开销下,MNOS可以有效降低SDN控制层被成功攻击的概率,并具备良好的容错/容侵能力;在此基础上,提出的选调策略和判决机制,可以有效提升系统的异构度和判决的准确性,进一步提升安全性能.

As a mission-critical network component in software defined networking(SDN),SDN control plane is suffering from the vulnerabilities exploited to launch malicious attacks,such as malicious applications attack,modifying flow rule attack,and so on.In this paper,we design and implement mimic network operating system(MNOS),an active defense architecture based on mimic security defense to deal with it.In addition to the SDN data plane and control plane,a mimic plane is introduced between them to manage and dynamically schedule heterogeneous SDN controllers.First,MNOS dynamically selects mcontrollers to be active to provide network service in parallel according to a certain scheduling strategy,and then judges whether controllers are in benign conditions via comparing the m responses from the controllers,and decides a most trusted response to send to switches so that the minority of malicious controllers will be tolerated.Theoretical analysis and experimental results demonstrate that MNOS can reduce the successful attack probability and significantly improve network security,and these benefits come at only modest cost:the latency is only about 9.47% lower.And simulation results prove that the scheduling strategy and decision fusion method proposed can increase system diversity and the accuracy of decisions respectively,which will enhance the security performance further.