In a previous paper, we presented a new method to generate evaluation data for the evaluation of security products and services. That approach tackles the issues of producing a workload with a rich semantic at a large scale. Testbed environments are the most appropriate tool for such task but induce a lot of effort and costs to implement. We presented a model to produce semantic data that can be implemented on light virtual networks and thus deployed at a large scale. This paper is an extension of our complete formal model. In this extension, we identify additional requirements for our model and define our ambitions. We translate those ambitions in verifiable properties of our model. Our prototype, although currently limited, provides the basis for an evaluation method that is customizable, reproducible, realistic, accurate and scalable. We generate realistic activity for up to 250 simulated users interacting with a real-world webmail server in an experiment to verify the properties of our mode