CERN is presently designing a new chain of accelerators to replace the present Proton Synchrotron (PS) complex: a 160 MeV room-temperature H- linac (Linac4) to replace the present 50 MeV proton linac injector, a 3.5 GeV Superconducting Proton Linac (SPL) to replace the 1.4 GeV PS booster (PSB) and a 50 GeV synchrotron (named PS2) to replace the 26 GeV PS. Linac4 has been funded and the civil engineering work started in October 2008, whilst the SPL is in an advanced stage of design. Beyond injecting into the future 50 GeV PS, the ultimate goal of the SPL is to generate a 4 MW beam for the production of intense neutrino beams. The radiation protection design is driven by the latter requirement. This thesis summarizes the radiation protection studies conducted for Linac4. FLUKA Monte Carlo simulations, complemented by analytical estimates, were performed 1) to evaluate the propagation of neutrons through the waveguide, ventilation and cable ducts placed along the accelerator, 2) to estimate the radiological impact of the accelerator in its low energy section, where the access area is located, and 3) to calculate the induced radioactivity in the air and in the components of the accelerator. The latter study is particularly important for maintenance interventions and final disposal of radioactive waste. Two possible layouts for the CCDTL section of the machine were considered in order to evaluate the feasibility, from the radiological standpoint, of replacing electromagnetic quadrupoles with permanent magnet quadrupoles with high content of cobalt. The present work provides complete information on beam loss assumptions, accelerator structure and duct and maze design, in order to make the present results of sufficiently general interest and provide guidelines for similar studies for intermediate energy proton accelerators. In addition to the Linac4-specific studies, this thesis also discusses FLUKA simulations performed to test the capability of the code, in a proton therapy application, in predicting induced radioactivity from intermediate energy protons. The thesis also reviews the analytical models mostly used for the calculation of neutron streaming through penetration traversing shielding barriers, discusses the FLUKA simulations performed to test the reliability of these models and, on the basis of the simulations, derives a universal expression that can be used to estimate the neutron transmission through a straight duct in direct view of the source, model missing so far in the literature.