Revealing anomalies in data usually suggest significant - also critical - actionable information in a wide variety of application domains. Anomaly detection can support dependability monitoring when traditional detection mechanisms e.g., based on event logs, probes and heartbeats, are considered inadequate or not applicable. On the other hand, checking the behavior of complex and dynamic system it is not trivial, since the notion of "normal" – and, consequently, anomalous - behavior is changing frequently according to the characteristics of such system. In such a context, performing anomaly detection calls for dedicate strategies and techniques that are not consolidated in the state-of-the-art. The paper expands the context, the challenges and the work done so far in association with our current research direction. The aim is to highlight the challenges and the future works that the PhD student tackled and will tackle in the next years.