Providing a safer communication in the network improves the network credibility since WSN is often deployed over hostile environments. Isolating the selfish nodes from the routing path is mandate to build a strong network. The node might become irresponsible due to their low energy level, high congestion rate, etc. Categorizing the selfish nodes and normal nodes makes the network more reliable while routing the data packets. Therefore an adaptive trust computation model is proposed here. By computing trust values with control messages and energy levels for the nodes are identified. This removes the selfish nodes from the system and the routes are constructed with the avoidance of irresponsible nodes. Here adaptive trust computation model is proposed with two level node selections. Simulation results are analyzed for determining the efficiency of the proposed scheme.