Attribute reduction is one of core research subjects in rough set theory. By means of analyzing some existing attribute reduction algorithms, it found that they cannot effectively or correctly get reduction results. An attribute reduction recursion algorithm based on attribute diminishing strategy was presented in this paper. The proposed attribute reduction algorithm firstly calculate dependency degree of every condition attribute; and then in turn subtract condition attributes with smaller dependency degree; subsequently, calculate dependency degree of remaining attributes set and decide whether the dependency degree is 1, if it is, the algorithm is recursively implemented; Finally, the all won attributes set were merged into the attribute reduction set and the core attributes were obtained. The proposed attribute reduction algorithm can not only fast calculate out all attribute reduction and core attributes, but also operates simply and has less computation, so that efficiency of the algorithm is improved. The experiment shows that the proposed attribute reduction algorithm can more effectively reduct decision table and have stronger practicability.