The present invention discloses a true random number generator, which adopts an SP structure. The true random number generator comprises: n physical random sources, n digital processing modules, a diffusion layer, an aliasing layer and a post-processing module. The n physical random sources are used to generate n physical signals; the n digital processing modules are used to respectively match andreceive the n physical signals, and convert the n physical signals into n digital signals; the diffusion layer is used to receive the n digital signals, perform an exclusive-OR operation on the n digital signals, and output n-channel signals, wherein each-channel signals comprise physical signals of at least three physical random sources; the aliasing layer is used to receive the n-channel signals, mix the n-channel signals to realize repeated mix among the n-channel signals, and output mix results; and the post-processing module is used to receive the mix results output by the aliasing layer, and eliminate the deviation and dependence of the mix results. Thus, according to the random number generator disclosed by the present invention, the robustness of the random number generator can beimproved.