Abstract-The large availability of smart portable devices and the growing interest in developing Internet of Things (IoT) oriented software components make several heterogeneous data available for analysis purposes. In the context of Crisis Management Systems, this means that people owning mobile devices when involved in natural disasters or terroristic attacks may be considered information sources as the classical ones, e.g., sensors or surveillance cameras. Including the information from the citizens in the situational analysis processes comes with two main issues that need to be addressed: i) the source could deliver wrong data (voluntarily or by mistake) that damage the integrity and the correctness of the analysis, and ii) a significant amount of heterogeneous data need to be selected, filtered and aggregated, to provide to the operator a real-time snapshot of the situation depicted using only credible and relevant information. In this paper, we define and implement a relevance labeling strategy able to process information coming from heterogeneous sources aimed at crisis situations and to provide to the human operator all the details he needs. We include provisions for detecting and removing redundancies and misleading data that can slow down or compromise the process and the a-posteriori analysis. The filtering strategy is last applied to events collected for the Secure! crisis management service-based system, showing its application to three scenarios related to real crisis situations happened in the last year.