In power system, state estimation is the basis of protecting and controlling the power grid. However, attackers can interfere the normal state estimating processes by strategically altering certain measurements through false data injection (FDI) attack. To address this problem, a typical solution is deploying a large amount of phasor measurement units (PMUs) at substations with necessary protection. However, the high expenses has limited the large-scale deployment of PMUs in reality. In this paper we present a methodology to ensure the power system state estimation through assessing risks and further providing protection schemes considering constrains of the grid operator. Firstly, we propose a risk assessment model composed of quantifying the importance-level of substations and computing the attack cost under a given security strategy. Furthermore, based on the model we introduce an optimal countermeasure synthesis approach which maximizes the attack cost and protects the most important substations. Simulations on standard IEEE test cases demonstrate the effectiveness and scalability of our framework. Our results indicate that protecting a small number of strategically chosen substations can significantly enhance the system security towards FDI attacks.