In this paper, we introduce a belief propagation based technique to combat the effects of primary user emulation attacks (PUEA) in Cognitive Radio (CR) Networks. Primary user emulation attacks have been identified as the most serious threat to CR security. In a PUEA, a malicious user emulates the characteristics of a primary user and transmits over idle channels. As a result, secondary users that want to use the channels are tricked into believing that they are occupied and avoid transmitting on those channels. This allows the malicious user to use the channels uncontested. To moderate the effects of PUEA, we propose a defence strategy based on belief propagation. In our solution, each secondary user examines the incoming signal and calculates the probability that it was transmitted from a primary user. These probabilities are known as beliefs. The beliefs at secondary users are reconciled to an agreed decision by comparison to a predefined threshold. The decision is made by a secondary user on whether it is believed that received transmission on a channel originated from a legitimate primary user or from a primary user emulation attacker.