Opportunistic spectrum access (OSA), in particular cognitive radio, is a new paradigm for addressing the scarcity of available frequency spectrum. Spectrum sensing is a key component of OSA, which provides the functionality of quickly searching for available wireless channels. In this talk, we will discuss a family of novel approaches in spectrum sensing, ranging from single-user-single-channel case to multi-user-multi-channel case. First, we apply the framework of quickest detection to spectrum sensing in order to reduce the detection delay while keep a reasonable false alarm rate. In order to improve the robustness, two quickest sensing algorithms are proposed, which waive of any necessity of prior information about signal and noise. Then, these prior-knowledge-free algorithms will be integrated with algorithms utilizing signal features. Second, we consider multiple collaborative users using random censored gossiping for quickest spectrum sensing. The broadcast probability, as a function of local likelihood ratio, is optimized using variational analysis. Finally, we study the case of multiple users and multiple channels, where each user recommends its favorite channels to neighbors in order to improve the spectrum access efficiency. The frameworks of interacting particles and epidemic propagation dynamics are applied to analyze the ergodicity and transience of the knowledge dissemination procedure. The works in the talk are supported or motivated by NSF projects CCF-0830451 and CNS-1247834.