Accurate data-driven models for epidemic spreading are key to the understanding of epidemic dynamics and to the design of suitable non-pharmaceutical interventions. In this talk, I will address epidemic diffusion at the urban scale, focusing on the role that publicly available data and empirical sociological findings play in our ability to model social interactions, in general within a city or inside a specific gathering place. I will present a urban social network model and a framework to simulate epidemic outbreaks that allows to formulate hypotheses and provides multi-level control of policy options, thus permitting to define probabilistic scenarios and to evaluate them in terms of diffusion of the contagion in the population and in the surrounding territory.