异构网络环境中的自适应在线故障检测

通过研究网络流量异常检测，提出一种新的基于自适应自回归(AAR)模型的在线故障检测算法．该算法只利用标准管理信息库，因此检测不依赖于特定产品类别，适用于异构网络环境．验证了流量信号的非平稳特性，并采用模拟获取的网络流量拟合AAR模型．由于不必将整个时间序列进行分片和单独拟合，算法可以直接处理获取的新数据，实现真正意义上的在线故障检测．利用时间平均方法消除了网络噪声的影响．在实验中，故障检测结果与预设的故障场景完全对应，进一步证明了该算法的准确性．

Adaptive Online Fault Detection in Heterogeneous Network Environment

A novel online fault detection algorithm based on adaptive auto-regressive (AAR) model is proposed focusing on the anomaly detection of network traffic. Since the standard information in management information base is used uniquely, the algorithm is independent of any special product families and applicable to heterogeneous network environment. The network traffic is validated to possess a non-stationary characteristic and the simulated network traffic is adopted to approximate the AAR model. The algorithm can analyze the new collected data directly without separating the whole time series into pieces and fitting them individually, so in the true sense it realizes the online fault detection. The influence of network noise can be eliminated by time average method. The fault detection results of experiment are consistent with the given fault scenarios, which verifies the accuracy of the algorithm.