Trust is one of the basic assumptions that we use in every day life. In ad hoc networks, routing protocols implement trust implicitly between the nodes of the network. Unfortunately, the implicit trust relations are not used by nodes whereas the attacks on ad hoc routing protocols precisely lead to the violation of at least one of these trust relations.In this paper, we demonstrate that a node can use these implicit trust relations to reason on neighbor behaviors to detect malicious nodes. We first formalize implicit trust relations of the AODV protocol. Then, we show that, given these relations, a node is able to reason on the actions performed by its neighbors and so deduce information about their knowledge. Finally, we discuss how these relations and the deduced information can be used to supervise the neighbor behavior and detect malicious nodes.