With the increasingly closer deployment of connected and automated vehicles (CAV), it is essential to study the different algorithmic stages involved in the design of active embedded systems. These stages are related to perception, decision, and action. As for the decision-making part which will allow generating maneuvers and optimal and safe trajectories, it is crucial to estimate the risk of the current situation according to the attributes of the scene key components obtained from the perception stage. This perception is shared into five main key components (obstacle, road, ego-vehicle, environment, and driver). In this paper, we have proposed to use and extend this perception modeling concept to build a more exhaustive and multi-modal risk ontology. This ontology shows firstly that the knowledge of risk is much more extended and complex than the simple use of a TTC indicator, and secondly that there is a very strong interdependence and interaction of the key components attributes.