With the introduction and development of cloud computing technology, major Internet companies gradually put their business services and huge amount of user data information in the cloud for processing and storage because it can itself realize distributed management of large amount of storage hardware, CPU, graphics card and other computer hardware resources using software, and provide users with hardware services that cannot be supported in the original computer system. However, some attack methods and approaches in traditional computer networks have also migrated to the remote network space, posing a threat to the resources of enterprises and individuals. At the same time, in order to hide their attack traces, some attackers are using anonymous networks to attack the cloud service platform. To address the costly and inefficient drawbacks of the Tor de-anonymization method, this paper focuses on the process of deploying a non-blind watermarking scheme called RAINBOW into the Tor network for watermarking association, mainly by (1) simulating the Tor network traffic in the cloud environment using mainflow traffic simulation tools; (2) studying and designing the RAINBOW non-blind traffic injection method to mark and identify the data traffic entering the Tor network, and then conduct an example analysis on the effect of mark identification; (3) evaluate the feasibility of the scheme for the Tor network data in the real environment. The experimental results verify the potential feasibility of this scheme for traffic correlation tracking in the Tor network.