Feasibility of One-Class-SVM for Anomaly Detection in Telecommunication Network

The growing number of unauthorized activities and various trends of networking technologies in telecommunication network have added heavy burdens to telecommunication performance management (PM) system. One-class-support vector machine (OCSVM) is introduced in this paper, to automatically detect network anomalies. Real telecommunication performance data are employed in this paper to investigate the feasibility of OCSVM for anomaly detection. Experiments with small and large data sets demonstrate that OCSVM can not only detect the anomalies correctly, but also fast in a short time. The promising performances show that OCSVM is efficiently enough to meet with the anomaly detection task in telecommunication network. Key-Words: Telecommunication, Anomaly detection, Performance management, OCSVM