改进SVM实现的无人集群网络入侵检测框架

An Improved SVM-based Framework for Unmanned SwarmNetwork Intrusion Detection

分布式无人集群之间的异构通信网络具有结构复杂、覆盖面积大、数据流量大等特征,而传统网络入侵检测框架数据吞吐能力不足,限制了无人集群的抗入侵、防失控能力的发展。本文提出了一种改进SVM实现的无人集群网络入侵检测框架,采用管线(pipeline)对支持向量机算法进行改进既满足了集成学习的需求,实现了网络入侵的检测和分类,同时也与分布式无人集成框架具有更好的适配性,并能够依赖Spark流式框架使其适应大规模通信数据的快速计算处理需求。通过实验对改进后的框架进行了性能评估,结果表明:该检测框架在准确性、实时性等多个方面均具备良好的性能,能够有效地支撑无人集群进行通信网络的入侵检测。

The heterogeneous communication network between distributed unmanned swarms possesses such characteristics as complex structure,large coverage area,and high data traffic.However,the insufficient data throughput of traditional network intrusion detection frameworks has limited the development of anti-intrusion and anti-loss-of-control capabilities of unmanned swarms.In light of this,this study proposes an improved network intrusion detection framework for unmanned clusters based on SVM,of which improvements of the support vector machine algorithm by adopting the pipeline can not only meet the requirements of ensemble learning,but also realize the detection and classification of network intrusion.The improvements of pipeline have better adaptability to the distributed unmanned integration framework,and it can rely on the Spark streaming framework to adapt to the rapid calculation and processing requirements of large-scale communication data.Through experimental evaluation of the improved framework,the results show that it possesses excellent performance in terms of accuracy,real-time response,and other aspects,effectively supporting the need for intrusion classification detection of large-scale data in unmanned swarms.