Abstract: For orthogonal frequency division multiplexing（OFDM） systems, a new spectrum sensing is proposed based on goodness of fit testing. Specifically, the sensing is established based on a distance between a known theoretical distribution and an empirical distribution of samples of cyclic prefix and its copy data in the OFDM system. Simulations show that the proposed sensing outperforms over the existing sensing based on autocorrelation of cyclic prefix. Moreover, in a noise uncertainty environment, performances of the sensing are better than energy detection in some cases.