On Accurate and Scalable Anomaly Detection in Next Generation Mobile Network

Abstract

This paper proposes an adaptive sampling strategy to address the accuracy and scalability issues of anomaly detection at high-speed backbone side of Next Generation Mobile Network (NGMN). The proposed sampling strategy is formulated based on the network traffic condition. It is constituted by two important functions namely the traffic identification and the sampling decision. While the former utilizes spectral analysis to identify the severity status of the traffic flows, the latter exploits both the flow status and flow size to compute the optimal sampling rate. In addition, a renormalization process is proposed to address the scalability issue in the network. Our analysis demonstrates that the proposed technique is efficient in providing adequate statistics for detecting anomaly traffic and scales well to the high speed traffic of NGMN.