Skyline query processing has recently received a lot of attention in database community. Lately, Akrivi Vlachou and D. Christos considered how to efficiently process subspace skyline queries in peer-to-peer networks,and proposed the concept of “extended skyline set” to reduce the volume of data transferred in the preprocessing phase for the first time. However, the experimental evaluation shows that this data structure is extremely limited in reducing the volume of data transferred in the preprocessing phase. Motivated by these facts, this paper proposes an efficient algorithm, i.e. TPAOSS (three-phase algorithm for optimizing skyline scalar), to reduce the volume of data transferred in the preprocessing phase. TPAOSS algorithm is based on the semantic relationship between full-space skylines and subspace skylines, and transfers the data through three phases. In the first phase, it only sends full-space skylines. In the second phase, it receives seed skylines. In the third phase, it exploits Bloom filter technology to obtain and send the replicated objects with seed skylines on the subspaces. Particularly, the paper presents two efficient strategies to reduce the volume of data transferred in the second phase. Furthermore, it presents detailed theoretical analyses and extensive experiments that demonstrate these algorithms are both efficient and effective.