With the assistance of global view in software-defined networks (SDNs), some switch nodes can be selected for traffic monitoring, which ensures the security of SDNs with a minimum cost. However, identifying the flow-based smallest subset of switches with the maximum coverage for node selection is always an NP-hard problem. To address this challenge, a two-stage parameter-rectifying-based selection (TSPRS) scheme is presented for efficient traffic monitoring in SDNs. In this novel scheme, a multi-timescale and multi-centrality layered virtual network (MMLVN) model is first built according to the different network structures over various time periods. By virtue of the MMLVN model, a tensor-based eigenvector multi-centrality computing (TEMC) method is employed at the first-stage selection to identify the smallest subset of switches with the maximum coverage for obtaining the subset of primary core nodes. To obtain more accurate parameters of real-time flows, a modified MMLVN model is further built by combining the results of a bandwidth cost-effective polling (BCeP) algorithm and the first-stage selection. Then, based on the modified MMLVN model, the TEMC method is reused to obtain the final optimization subset of the core nodes for efficient traffic monitoring at the second-stage selection. Extensive simulation experiments demonstrate that the proposed scheme can achieve a desirable tradeoff between the monitoring overhead and accuracy, and also reduce the communication bandwidth cost, which is superior to the existing state-of-the-art schemes.

借助软件定义网络（SDN）的全局视图，可以选择一些交换节点进行流量监控，以最小的成本确保SDN的安全。然而，识别具有最大节点选择覆盖范围的基于流的最小交换机子集始终是一个 NP-hard 问题。为了应对这一挑战，提出了一种基于参数整流的两阶段选择 (TSPRS) 方案，用于 SDN 中的高效流量监控。在这个新颖的方案中，首先根据不同时间段的不同网络结构构建了多时间尺度和多中心分层虚拟网络（MMLVN）模型。凭借 MMLVN 模型，在第一阶段选择中采用基于张量的特征向量多中心计算（TEMC）方法来识别具有最大覆盖范围的最小交换机子集，以获得主核心节点子集。为了获得更准确的实时流参数，结合带宽成本有效轮询（BCeP）算法和第一阶段选择的结果，进一步构建了改进的 MMLVN 模型。然后，基于修改后的MMLVN模型，重用TEMC方法得到核心节点的最终优化子集，用于第二阶段选择时的高效流量监控。大量的仿真实验表明，所提出的方案可以在监控开销和精度之间实现理想的折衷，同时还降低了通信带宽成本，