The effects of input random variables’ uncertainties on failure probability can be characterized by failure probability based global sensitivity index, and they can provide guidance on how to reduce the failure probability of a structure. Based on support vector machine (SVM) method and monte-carlo (MCS) method, a new method is proposed for structural global sensitivity analysis in this study. Approximation model based on SVM is constructed to simulate the mapping relationship between input variables and output response, effectively overcome the issues of implicit limit state function and high computational effort. The uniform mapping, meanwhile, is employed to obtain a certain number of failure samples for a higher precision of approximation model. The approximate model based on SVM, under the small sample has good generalization ability, can significantly improve the computational efficiency, reducing the amount of calculation. Two numerical and an engineering examples are used to illustrate the accuracy and efficiency of the proposed method.