We address the problem of spectrum allocation by a moderator employed in a cognitive radio (CR) network in the presence of multiple primary users (PUs). It is assumed that the moderator is uncertain about the state of the PU channel (busy or vacant) and therefore, does not have perfect knowledge about spectrum availability. In order to improve the performance, secondary users (SUs) send their sensing decisions to the moderator. The moderator fuses the data from these SUs and makes an inference about spectrum availability. We assume that the spectrum allocation is made by the moderator only when it decides that spectrum is available. In this paper, we model the interaction between the PUs, SUs and the moderator within a bilateral trading framework, and present an iterative allocation mechanism that guarantees improvement in the revenue of the moderator in every iteration.