Abstract. Collaborative WSN are known to efficiently correlate sen- sors’ events to detect spatially-correlated events with a low latency. How- ever, because of the drastic reduction of messages sent to the network administrator, it is difficult for him/her to understand in what state the network is. In this paper, we propose a modality-agnostic collaborative detection of spatio-temporally correlated events that drastically lowers power consumption by both reducing and localising communication. This contribution can then be used to expose better auditing capabilities that overcome the problems usually found in collaborative networks. These ca- pabilities can in turn be used by both the administrator and the network itself to, for instance, automatically react to faulty sensors. Experiments validate the interest of our proposal in an intrusion detection scenario.