A cyber-physical model for wide-area control of power systems is considered, where the state variables of each generator are measured and sent to the cyber-network and the corresponding control inputs are computed distributively. The secure control of such wide-area power systems is considered in the presence of cyber attacks that introduce threats that compromise their integrity and confidentiality. Detection, prevention, and resilience for these attacks and algorithms for accomplishing these goals are proposed. In particular, an algorithm to overcome confidentiality attacks of the underlying control gains is presented. Also proposed is an algorithm for defense against integrity attacks that might take place on the cyber-network. For this purpose, a resilient information retrieval approach is leveraged which recovers the true state variables despite the malicious attacks on both virtual machines and communication links. The retrieved states are then used to detect possible attacks on phasor measurement units (PMU) in the next time-step. Simulation studies are included to validate our proposed approaches.