Over the last decades a number of guidelines have been proposed for best practices, frameworks, and cyber risk assessment in present computational environments. In order to improve cyber security vulnerability, in this work it is proposed and characterized a feasible methodology for problem solving that allows for the evaluation of cyber security in terms of an estimation of its entropic state, i.e., a predictive evaluation of its risk and vulnerabilities, or in other words, the cyber security level of such ecosystem. The analysis and development of such a model is based on a line of logical formalisms for Knowledge Representation and Reasoning, consistent with an Artificial Neural Networks approach to computing, a model that considers the cause behind the action.