In this paper, we investigate the challenge in passive localization via asynchronous TOA (time of arrival) measurements of a signal source in wireless sensor networks (WSN). Due to the lack of perfect time synchronization between the anchor nodes and the signal source node, the unknown parameter of start transmission time of signal source further complicates the localization problem. The derived maximum likelihood estimator cost function is nonlinear and nonconvex with multiple local minimum. To solve the global minimum, a novel two-step method is implemented. First, we creatively proposed an improved Monte Carlo (MC) method, named the maximum (minimum) distance maximum (minimum) TOA matching based (MDTM-based) second-order MC method, by taking into account the relationship between the TOA measurements and the actual distance between the anchor nodes and source node. We use the proposed method to find a rough initial position of the signal source with low computational complexity by leveraging dimensionality reduction. Then, the rough initial position value is refined by using trust region (TR) method to obtain the final positioning result. Simulation comparatively demonstrates that the proposed method outperforms the state-of-the-art methods in terms of localization accuracy and computational complexity.