Methods and systems are disclosed for processing policies. A query policy tree with expressions and combination rules is received. A second policy tree stored in a policy repository is searched to select a sub-policy in a neutral format, reflecting a structure with a combination rule and a leaf with a condition expression. A depth of the structure is reduced through removal of redundant combination rules and combinable leaf expressions. The sub-policy tree is selected from the second stored policy tree by comparing the query policy tree. The selected sub-policy tree is copied to a third policy tree. Leaves of the third policy tree are evaluated against the leaves of the query policy. The nodes of the third policy are joined against the node of the query policy. The third policy is reduced to the neutral format by combining evaluated leaves and removing redundant combination rules. Finally, a response is provided.