In traditional cognitive radio (CR) network, most existing graph-based spectrum allocation schemes don’t take on-off behavior of primary users (PUs) into consideration. In this paper, a novel spectrum allocation algorithm based on the activities of the PUs is proposed. The proposed algorithm mainly focuses on the vacant probability of licensed spectrums. And it allocates the vacant spectrums considering the interference to the neighbor cognitive nodes and the probability fairness of different cognitive nodes during the allocation. Based on the definition of the obtained benefit of cognitive node, new utility functions are formulated to characterize the system total spectrum utilization and fairness performance from the perspective of available probability. The simulation results validate that the proposed algorithm with low system communication cost is more effective than the traditional schemes when the available licensed spectrums are not sufficient, which is effective and meaningful to a real CR system with bad network condition. In this paper, a novel spectrum allocation algorithm based on the activities of the PUs. In this paper, a novel spectrum allocation algorithm based on the activities of the PUs is proposed. The proposed algorithm mainly focuses on the vacant probability of licensed spectrums. And it allocates the vacant spectrums considering the interference to the neighbor cognitive nodes and the probability fairness of different cognitive nodes during the allocation. Based on the definition of the obtained benefit of cognitive node, new utility functions are formulated to characterize the system total spectrum utilization and fairness performance from the perspective of available probability. The simulation results valid that that proposed algorithm with low system communication cost is more effective than the traditional schemes when the may licensed spectrums are not sufficient, which is effective and meaningful to a real CR system with bad network condition.