High traffic load over the unlicensed portion of the radiospectrum (a.k.a., ISM bands) along with inefficient usage of thelicensed spectrum gave impetus for a new paradigm in spectrumallocation, whose main purpose is to improve spectrum efficiencythrough opportunistic access. Cognitive radios (CRs) havebeen proposed as a key enabling technology for such paradigm.Operating a CR network (CRN) without impacting the performance oflicensed (primary) users requires new protocols for informationexchange as well as mathematical tools to optimize thecontrollable parameters of the CRN. In this dissertation, wetarget the design of such protocols. First, we develop adistributed CRN MAC (COMAC) protocol that enables unlicensed usersto dynamically utilize the spectrum while limiting theinterference they inflict on primary (PR) users. The main noveltyin COMAC lies in not assuming a predefined CR-to-PR power mask andnot requiring coordination with PR users. Second, we propose anovel distance-dependent MAC protocol for CRNs in whicheach CR is equipped with multiple transceivers. Our protocol(called DDMAC) attempts to maximize the CRN throughput byfollowing a novel probabilistic channel assignment mechanism. Thismechanism exploits the dependence between the signal's attenuationmodel and the transmission distance while considering the trafficprofile. We show that through its distance- and traffic-aware,DDMAC significantly improves network throughput. Finally, weaddress the problem of assigning channels to CR transmissions,assuming one transceiver per CR. The main goal of our design is tomaximize the CRN throughput with respect to both spectrumassignment and transmission power. Specifically, we presentcentralized and distributed solutions that leverage the uniquecapabilities of CRs. Compared with previously proposed protocols,our schemes are shown to significantly improve network throughput.