Multiple location-dependent applications of wireless sensor networks have been proposed, such as wildlife monitoring and instruction detection. A reliable and low-cost localization scheme is essential for such sensor applications. In this paper, we propose a signal-strength-based cooperative localization algorithm. Any wireless node monitors all other neighbor nodes' signals, not only from beacon nodes but also from unknown nodes. The signal strength from beacons can be used to limit the position region of the unknown node, and the signal strength from other unknown nodes is iteratively used to refine the node's position by applying greedy method. This scheme does not need labored prior training. Simulation results show that the localization precision is satisfied, and the localization error is reduced greatly compared with the traditional beacon-only-based algorithm