This paper presents a double-threshold cooperative spectrum sensing algorithm which is based on trust and satisfies both reliability and efficiency. The cognitive nodes that satisfy the request of double-threshold have the priority to participate in cooperative sensing and that satisfy the requirement of trust parameters may participate in cooperative sensing if only the number of the former is smaller than a preset value. The fusion center stores the sensing record of each cognitive node and sets the fusion weights according to the partial detected results. Theoretical analysis and simulation show that the bandwidth required for transmitting the sensing parameters decreases, and the detection performance improves because the unreliable users are reduced. Additionally, the algorithm can be made to adapt to different wireless service by adjusting the parameter nt.