The objective of the proposed research is to develop "interventional security" for mitigating effects of cyber-physical attacks against power-control networks. The goal is to develop a real-time, autonomous decision layer (and supporting components) that fuses power-grid, communications-network, and field-device information for providing continuous cyber-physical security-state estimation, determining appropriate system-wide and device-specific security policies, alerting on abnormal or suspicious conditions, and approving control actions automatically or requesting approval by a trusted operator. The hypothesis herein is that it is possible to design a system that can discern questionable control actions and cyber-physical trajectories, with development focused on how these decisions are made and when to intervene. Due to the complex and disjoint nature of power systems and communications networks, simple rule-based solutions are insufficient. The core contribution and innovation lies in development of a method sophisticated enough to deconflict and to harmonize disparate information using formal, statistical-inference-based methods.