Although recursive localization approaches stand well in small-scale networks, they are unsuitable for largescale WSNs. Indeed, they suffer from the adverse effects of error propagation and accumulation. They also require an important communication overhead related to the localization data broadcasted by new reference nodes. To deal with these issues, we developed a new reliable reference selection strategy that ensures a better distribution of these nodes. Our approach couples the novel selection strategy with a refinement phase that enhances the position accuracy.