In this work, we propose GST-CEMID, a pipelining-based framework to support Complex Event ModelIng and Detection in MSNs, which relies on: (i) an extension of GStreamer; (ii) JSON-like configuration files; and (iii) an ontology-based data model. GST-CEMID allows users to model MSN infrastructure and complex events, and generates an event processing pipeline according to the provided models. Results show that it can detect events in a high workload with the detection latency of less than one second.