The paper proposes a 5-layered context model intended as a mean to drive context-aware systems. An ontology based model (CONON) is defined, which facilitates explicit context representation, semantic knowledge sharing and logic reasoning. A temporal ontology is included in the model to represent temporal context information and sample inferences are provided to demonstrate how implicit context can be derived from low level explicit context. The proposed architecture is deployed in 5 layers: acquisition layer, representation layer, aggregation layer, interpretation layer and utilization layer. The acquisition layer converts sensor data into semantic information in the representation layer. Such low level information is integrated by means of a centralized knowledge repository where information coming from different sensors is stored and enriched by relations. The context interpretation layer leverages reasoning rules and/or learning mechanisms to deduce high-levek implicit knowledge from low level explicit information. Eventually the utilization layer hosts context-aware services that use both low-level and high-level information to customize their behavior according to the current situation. The proposed context model is applied on a context-aware smart home implemented as an OSGi residential gateway.