A Novel Network Traffic Anomaly Detection Approach Using the Optimal $\varphi$-DTW

Under the current severe situation of cyber security, it is of great significance to propose an effective anomaly detection approach for ensuring the stability of network. It is generally known that the network traffic data is a kind of typical streaming time series data, which are recorded by network equipments usually accompanied by time instants. In order to detect the anomalous sections in network traffic data effectively, we propose an unsupervised anomaly detection approach based on anomaly definition in time series by utilizing the optimal φ-DTW and the corresponding similarity matrix, which is called ADOPD. Comprehensive experiments have demonstrated that our proposed approach achieves satisfying performance on detecting anomalous in real world data sets.