This paper addresses the vulnerability of the power system subjected to physical deliberate attacks (PDAs) when the power system and communication network are geographic-cyber interdependent. This interdependence is due to the use of the transmission line towers as an available infrastructure to establish the communication network that controls the power system. The attacker plans a PDA on both transmission lines and the communication network to damage the power system, and in reverse the system operator (SO) reacts as a defender by taking necessary actions against them. To this end, a bilevel optimization based model is presented. The upper level problem formulates the attacker's problem with the objective function of maximizing damage inflicted on the power system through the simultaneous PDA on the transmission lines and communication links. In the lower level, the SO's problem is modeled with the objective function of minimizing damage to the system through the corrective actions including the change in the production level of generation units and the load shedding. The SO's corrective actions are performed through a main intact connected communication network and the alternative infrastructure according to the power system operation strategy in emergency. For the proposed bilevel model, a solution approach is proposed based upon the combination of the genetic algorithm and mixed integer linear programming. Finally, numerical results obtained by implementing the model on the 24-bus IEEE reliability test system are presented and then compared with previous studies. The efficacy of the proposed model is confirmed by different numerical case studies.