In this paper we consider the problem of exploiting spectrum resources within the Opportunistic Spectrum Access context. We mainly focus on the case where one secondary user (SU) probes a pool of possibly available channels dedicated to a primary network. The SU is assumed to have imperfect sensing abilities. We, first, model the problem as a Multi-Armed Bandit problem with sensing errors. Then, we suggest to analyze the performances of the well known Upper Confidence Bound algorithm UCB1 within this framework, and show that we still can obtain an order optimal channel selection behavior. Finally we compare these results to those obtained in the case of perfect sensing. Simulation results are provided to support the suggested approach.