Although collaborative spectrum sensing can greatly improve sensing performance, transmission efficiency
will decrease with the number of sensing nodes that consume more system resources. The sensing performance of
collaborative spectrum sensing is investigated, and relationship among the performance, node number and average
signal-to-noise ratio derived. Based on the study, a method for selecting a set of sensing nodes is proposed. A
method for determination of the confidence degree is presented to eliminate trustless nodes in the cognitive radio
(CR) network, which adds negative effect to the selection method. Using the above two methods, an algorithm for
selecting sensing node set based on confidence is proposed, which can effectively eliminate trustless nodes in the CR
network and guarantee the sensing performance of the set. Simulation results show validity and reliability of the
algorithm.