Abstract: The queue stability of the secondary users in cognitive network was considered. A Markov chain model for the state of users in the cognitive network was constructed and a distributed CSMA algorithm was proposed. The secondary users adjust the parameters of back-off time due to the arrival and service rate to ensure the queue stability. Upper bound of the capacity of secondary users under the collision constrainting to primary user was derived in closed form. If the arrival rate of secondary users is smaller than the upper bound of capacity, queue stability can be ensured by the proposed algorithm. Simulations verified the effectiveness of the algorithm.