A BRPCA based approach for anomaly detection in mobile networks

Abstract

Researchers have recently uncovered numerous exploitable vulnerabilities that enable malicious individuals to mount attacks against mobile network users and services. The detection and attribution of these threats are of major importance to the mobile operators. Therefore, this paper presents a novel approach for anomaly detection in 3G/4G mobile networks based on Bayesian Robust Principal Component Analysis (BRPCA), which enables cognition in mobile networks through the ability to perceive threats and to act in order to mitigate their effects. BRPCA is used to model aggregate network data and subsequently identify abnormal network states. A major difference with previous work is that this method takes into account the spatio-temporal nature of the mobile network traffic, to reveal encoded periodic characteristics, which has the potential to reduce false positive rate. Furthermore, the BRPCA method is unsupervised and does not raise privacy issues due to the nature of the raw data. The effectiveness of the approach was evaluated against three other methods on two synthetic datasets for a large mobile network, and the results show that BRPCA provides both higher detection rate and lower computational overhead.