Self-pierce riveting (SPR) is one of the commonly used joining technologies, which is useful to assemble dissimilar materials. However, the galvanic / crevice corrosion between joining metals and metal-rivet interfaces plays an important role on the mechanical properties of the joint. It is critical to have a compressive understanding on the corrosion phenomena of the joint. In this study, a hybrid model that combines the stochastic corrosion nucleation method and physics-based finite element (FE) modeling is proposed. The corrosion nucleation variables, including the positions of the nucleation sites and their corrosion trigger time, are firstly obtained based upon statistical analysis on experiments and then imported in the physics-based FE model as initial conditions. Afterwards, the corrosion propagation process is evaluated via the FE model. Various corrosion scenarios with different nucleation variables are generated and statistically analyzed. Experimental results are used to validate the model. It is found that, the proposed hybrid model can reasonably predict the localized corrosion results of the SPR joints, compared to experimental observations. Moreover, the corrosion area and material loss in the SPR joint are predicted.