A group key management is essential method guaranteeing of data confidentiality and integrity in secure multicast communication. For secure multicast, group key management schemes have focused on communication cost, i.e., the number of update messages, to efficiently use bandwidth. To reduce the communication cost, many group key management schemes adopt a key tree structure. But, to adopt a key tree structure, a more powerful capability is needed, as a tradeoff, to store various kinds of keys and to execute several cryptography algorithms, i.e., storage cost and computation cost, respectively. But, it is important to simultaneously reduce not only the overhead of communication cost but also the overheads of storage and computation costs in mobile devices which have limited storage space and computation capability. In this paper, we propose a key tree structure, and a key tree management protocol for the efficiency of storage and computation costs. The proposed protocol adaptively controls the key tree structure not only to reduce the increase of the communication cost and but also to minimize the storage and the computation costs by using the limited resource information from each group members. We analyze the proposed protocol with three kinds of cost metrics and conduct simulations. The results show that the proposed protocol can save storage and computation costs. As a tradeoff, the proposed protocol requires a very small increment of communication cost if the ratio of leaving group members and the number of group members are moderately large in a batch rekeying interval.