Accurate spectrum sensing is very important for cognitive radio systems. In the cooperative spectrum sensing, the unreliable nodes will result in deterioration of detection performance. This paper focuses on the spectrum sensing in the case of unreliable nodes and proposes a weighted reliability cooperative spectrum sensing algorithm. In the algorithm, the global fusion center only chooses the cognitive nodes whose reliabilities are larger than the average one. The simulations and results show that the algorithm can improve the spectrum sensing performance effectively.