Dynamic spectrum auction has been widely recognized as a promising solution to spectrum allocation in cognitive radio networks. However, the performance of dynamic spectrum auction is significantly affected by the uncertainty of the available time of spectrum opportunities. To combat this uncertainty, we propose risk-reduced auction (RRA) mechanism in this paper. Specifically, we study the scenario where a cognitive base station (CBS) auctions spectrum opportunities to secondary users (SUs). Modeling the traffic pattern of primary users (PUs) as an alternating renewal process, we derive the optimal auction time to maximize the auctioneer's utility to combat the time-duration uncertainty. In the meantime, a collision probability constraint is imposed to protect the PU's priority on spectrum utilization. We show analytically that both the auctioneer and the SUs are truthful and there is a weakly dominant equilibrium in the RRA algorithm. Moreover, the SUs bidding risk can be effectively reduced. Also conducted is a set of simulation evaluations which demonstrate the improvement of both auctioneers and SUs' utilities achieved by using our proposed RRA algorithm.