This letter presents a low complexity cyclic autocorrelation function (CAF) computation technique for the spectrum sensing of orthogonal frequency division multiplexing (OFDM) signals in cognitive radio. Although CAFs are frequently employed in signal cyclostationarity detection (CD) based spectrum sensing, the computational complexity of CAFs is greater than that of signal energy. The presented technique computes the CAF by multiplying a received signal, delayed one and complex periodic function where these signals and function are thinned-out so that the sampling frequency becomes quadruple of a cyclic frequency. The employed function can only take the values ±1 and 0. Because the multiplication and addition can omit in the CAF computation, the computational complexity can be reduced. We numerically verify the effectiveness of the presented CAF computation technique.