VoIP infrastructures are exposed to a large variety of security attacks, but the deployment of security safeguards may deteriorate their performance. Risk management provides new perspectives for addressing this issue. Risk models permit to reduce these attacks while maintaining the quality of such a critical service. These models often suffer from their complexity due to the high number of parameters to be configured. We therefore propose in this paper a self-configuration strategy for support- ing runtime risk management in VoIP architectures. This strategy aims at automatically adapting these parameters based on an econometric feedback mechanism. We mathematically describe this self-configuration strategy, show how it can be integrated into our runtime risk model. We then evaluate its deployment based on a proof-of-concept prototype, and quantify its performance through an extensive set of simulation results.