Host identification is very important for computer forensics and anonymous attack resistance. In order to accurately identify the target host on the network, the definition and properties of the multi-dimensional host fingerprint model are given and formalized. Then, in view of the problem of reliability and accuracy of fingerprint acquisition, this paper proposes a multi-dimensional host fingerprint model for high-speed hybrid network traffic, which integrates the hardware characteristic information, host software environment characteristic information and host network behavior characteristic information. The experimental results show that the proposed model can extract data flexibly and efficiently in the high-speed hybrid network, and the multi-dimensional host fingerprint model can effectively identify the host with the accuracy of 93.33%, which has increased by nearly 8 percent compared with the single-dimension host fingerprint identification, and the multi-dimensional host fingerprint model is not affected by IP address changes. In general, the multi-dimensional host fingerprint model has higher reliability and accuracy compared with the single-dimensional host fingerprint identification.