A Feature Extraction Method of Network Traffic for Time-Frequency Synchronization Applications

Network traffic embodies network behaviors and users' activity patterns, which holds many inherent features and dynamic properties. Feature extraction for network traffic in the network plays an significantly important role in time-frequency synchronization applications. How to accurately extract the hidden properties and features of network traffic has an important impact on network activities, such as network failure positioning, anomaly detection, and performance analysis. To this end, this paper propose a new feature extraction method to characterize network traffic. Firstly, the time-frequency analysis theory is used to transform network traffic in time-frequency synchronization applications to the time-frequency domain. Then the cluster analysis theory is used to dig and extract network traffic feature components. And the k-means analysis method is exploited to refine the hidden features of network traffic in the time domain. Finally, to validate our feature analysis method, we conduct an anomaly detection test. Simulation results show that our approach is promising.