In this paper, we address an important issue of trust evaluation for tactical wireless sensor networks, where energy-hungry sensor nodes are deployed in harsh and hostile areas for the purpose of surveillance and reconnaissance. In particular, due to the limited resources of the node, it is of great importance to detect and isolate malicious nodes in an energy-efficient manner. Most of previous work adopt some proactive scanning method, requiring periodical exchange of control messages even in ordinary times and thus significant computational overhead for updating trust values per all nodes over the network. In order to provide more energy efficient evaluation method (while yet preventing any intelligent attacks such as selective forwarding and false positive problem attacks), we propose a novel cooperative trust evaluation scheme in which the initial trust level of nodes measured by ondemand trust rating approach is re-evaluated later on by a root or gateway node based on a pattern analysis. Through a preliminary simulation study using OPNET simulator, we compare our scheme to the existing work and prove that it can achieve the higher detection rate with much less energy consumption.