In this paper we address the node localization problem in large-scale wireless sensor networks (WSNs) by using the received signal strength (RSS) measurements. According to the conventional path loss model, we first pose the maximum likelihood (ML) problem. The ML-based solutions are of particular importance due to their asymptotically optimal performance (for large enough data records). However, the ML problem is highly non-linear and non-convex, which makes the search for the globally optimal solution difficult. To overcome the non-linearity and the non-convexity of the objective function, we propose an efficient second-order cone programming (SOCP) relaxation, which solves the node localization problem in a completely distributed manner. We investigate both synchronous and asynchronous node communication cases. Computer simulations show that the proposed approach works well in various scenarios, and efficiently solves the localization problem. Moreover, simulation results show that the performance of the proposed approach does not deteriorate when synchronous node communication is not feasible.