As a basic dynamical process, random walk on networks is fundamental to many branches of science, and has attracted much attention. A difficult problem in the study of random walk is how to obtain the exact solution for the mean trapping time (MTT) of this process. The MTT is defined as the mean time for the walker staring from any node in the network to first reach the trap node. In this paper, we study random walk on the Koch network with a trap located at the highest degree node and calculate the solution for MTT. The accurate expression for the MTT is obtained through the recurrence relation and the structure properties of the Koch network. We confirm the correctness of the MTT result by direct numerical calculations based on the Laplacian matrix of Koch network. It can be seen from the obtained results that in the large limit of network size, the MTT increases linearly with the size of network increasing. Comparison between the MTT result of the Koch network with that of the other networks, such as complete graph, regular lattices, Sierpinski fractals, and T-graph, shows that the Koch has a high transmission efficiency.