Managing the configuration of heterogeneous enterprise security mechanisms is a complex task. The effectiveness of a configuration may be constrained by poor understanding and/or management of the overall security policy requirements, which may, in turn, unnecessarily expose the enterprise to known threats. This paper proposes a threat management based approach, whereby knowledge about the effectiveness of mitigating countermeasures is used to guide the autonomic configuration of security mechanisms. This knowledge is modeled in terms of Semantic Threat Graphs, a variation of the traditional Threat/Attack Tree, extended in order to relate semantic information about security configuration with threats, vulnerabilities and countermeasures. An ontology-based approach to representing and reasoning over this knowledge is taken. A case study based on Network Access Controls demonstrates how threats can be analysed and how automated configuration recommendations can be made based on catalogues of countermeasures. These countermeasures are drawn from best-practice standards, including NIST, IETF and PCI-DSS recommendations for firewall configuration.