The forthcoming 5G operational environment entails heterogeneous and multi-dimensional ecosystems where cyber assets, digital actors and cyber-physical risks coexist. In this context, the prediction and anticipation of the attacks propagation thorough the targeted systems promises to be some of the major workhorses of the emerging self-protection capabilities. In the grounds of the Self-Organizing Network (SON) paradigm, it is expected that by taking into account proactive actuations, the decision and enforcement of the best courses of action will be enhanced. With the aim on contributing to their planning and execution, this paper introduces a novel framework for proactive self-protection on 5G environments, the description of an architectural framework able to sustain the rest of the anticipation enablers, the formalization of a knowledge representation and reasoning strategy for active cyber threat mitigation, and a prediction strategy adapted to the difficulties inherent in analyzing events on 5G scenarios. The effectiveness of the proposal has been demonstrated by proof-of-concept instantiation for anticipating the impact of Denial of Service (DoS) attacks on a real communication environment.