An advanced response surface method based on Kriging theory, named as K-RSM, is proposed in this paper to assess the stability of tunnel face driven in soils. An efficient uniform design is employed to obtain a set of sampling points that can exactly approach the limit state surface for establishment of K-RSM. By designing a series of initial sampling points that are uniformly distributed in the space of random variables, the bisection method is adopted to search the points that make limit state function equal to 0. A Monte Carlo population with a reasonable size is generated and predicted by the trained K-RSM to estimate the failure probability of tunnel face. Two typical examples, involving closed response surface and multi-dimensional nonlinear problem respectively, are discussed to test the efficiency and accuracy of the proposed K-RSM. Probabilistic stability analysis of three-dimensional (3D) tunnel face is performed using K-RSM. Results show that the proposed K-RSM is an accurate and time-saving approach to implement reliability analysis of tunnel face, especially for cases with low failure probability.