Protection of smart grids against cyber attacks has become one of the nation's top priorities. Most existing approaches have focused on intrusion detection and post-attack responses in a similar fashion to fault detection and recovery. However, attack behaviors are much more complex and less predictable than fault behaviors as an attacker often changes his course of actions during the attacks. Thus, there is a need for a systematic approach that can incorporate behaviors of attackers into the decision analysis of security management. This is especially crucial as attackers can keep advancing their techniques to outsmart security technologies. Our research aims to enhance the science of security that enables behavioral modeling as well as model extensibility. In particular, we present an analytical game theoretic approach to analyzing security of smart grid SCADA (Supervisory Control and Data Acquisition) systems by including attacker/defender behaviors in the proposed sequential, non-zero sum, and two-player game model. The paper describes the development of the game payoffs and illustrates how informed decisions can be made on a real-world scenario of attacks at the sensor level of the smart grid SCADA systems.