In this paper, we introduce a new methodology based on Complex Network Theory aimed at the prompt identification of leakages occurring in a water distribution system. The underlying idea is a data mining algorithm which, starting from real-time measurements of pressure gages installed in the system, builds a virtual complex network in which the nodes are represented by the locations of instrumentation and links are created if the correlation coefficient between pressure signals is above a given threshold. For the system analyzed in the paper, we show that the node in which the leak forms is always characterized by the lowest degree centrality with respect to all the others, thus giving the possibility of its early identification.
