In this work we present a novel scheme for statistical-based anomaly detection in 3G cellular networks. The traffic data collected by a passive monitoring system are reduced to a set of per-mobile user counters, from which time-series of unidimensio