Abstract: This papar presents a probabilistic logging scheme based on Bloom filter for source tracing. The scheme makes probabilistic sampling of all packets through each router, and uses efficient Bloom filter for storage. The sampling information can stored in memory, which make it easier to find. This paper introduces first the concept of source locating server. Besides forwarding packets, the routers in the core network only need probabilistic sampling of packets. In addition, this paper gives theoretical analysis of the choice of the relevant parameters. In theory, This paper analyzes the cost of storage in probabilistic logging scheme and the validity of source location. The proposed scheme has the characteristics of small storage costs and high efficiency, which provides a theoretical basis for further actually deplyment.