Much of the envisioned usefulness of wireless sensor networks results directly from their ad-hoc deployment and self-organization ability in critical and adverse environments, where they are to provide timely and reliable information constantly or on demand. In such environments, however, networks become highly susceptible to attacks at every layer of the network stack. But the extreme resource starvation of nodes poses severe challenges for security implementations. That's why most of the security approaches for general ad-hoc networks don't work for sensor networks; trust based approaches seem most likely suitable, but work best in mobile environments. In this work we are extending the trust paradigm for sensor networks to consider the prevailing mode of communication and topology in the network. By introducing hidden probing messages malicious nodes are confronted with a forwarding dilemma, which can be considered as a repeated game of trust. We show that malicious nodes that don't cooperate lose trust in their downstream neighbors, connectivity and thereby criticality to the network.