Access to electric power is critical to societal welfare. In this paper, we analyze the interaction between a defender and a terrorist who threatens the operation of an electric power system. The defender wants to find a strategic defense to minimize the consequences of an attack. Both parties have limited budgets and behave in their own self-interest. The problem is formulated as a multi-level mixed-integer programming problem. A Tabu Search with an embedded greedy algorithm for the attack problem is implemented to find the optimum defense strategy. We apply the algorithm to a 24-bus network for a combination of four different defense budgets, three attack budgets, and three assumptions as to how the terrorists craft their attacks.