基于时频分析的网络故障自动识别

通过研究网络流量的时频分析,提出了一种新的基于平滑魏格纳分布(WVD)的故障识别算法.该算法只利用管理信息库中的标准信息来识别不同种类的故障,而且与现有的简单网络管理协议体系结构兼容.采用平滑WVD可消除交叉项的干扰,并将网络流量序列转换为二维空间的波动能量分布.用获取的不同网络服务的时频特性分布作为训练样本,训练后的K最近邻分类器可实现网络故障的识别.实验中,故障识别结果与预设的场景一致,与理论值相比识别误差率为14 41%.分析结果表明,该算法适用于具有流量变化的故障场景.

Automatic Identification of Network Fault Based on Time-Frequency Analysis

A novel fault identification algorithm based on smooth pseudo Wigner-Ville distribution (WVD) is proposed, which focused on the time-frequency analysis of network traffic. Since the algorithm merely utilizes standard information in management information base to identify the heterogeneous faults, it is compatible with the existing simple network management protocol framework. After adopting the smooth pseudo WVD that reduces the interference of the cross term, the series is transformed into a two-dimension distribution of fluctuation energy. Training samples are acquired from the time-frequency characteristic distribution of different network services. A K-nearest neighbor classifier is used to identify the system faults after being trained. The fault identification result of experiment is consistent with the given scenarios,which proves the validity of the algorithm. Compared to the theoretic value, the identification error ratio is 14.41%. The corresponding analyses further prove that the novel fault identification algorithm is (adaptive) to the fault scenario with changing of network traffic.