This paper presents a bottom-up approach to implement the automatic and scientific transform of access control policies in the migration. First, the problem of mining security labels optimally is described formally, and it is then proved to be NP-complete. Next, an approximate optimization algorithm based on hierarchical clustering and genetic algorithm is presented, which decomposes the problem into two parts: category partition and secret level assignation. Finally, experimental results show that the algorithm is effective in finding an optimal solution. The proposed approach can be applied to migration projects in hierarchy protection in information security.