In this talk we propose three GNU Radio blocks for performing spectrum sensing based on the autocorrelation of samples captured with an SDR device such as HackRF One, RTL-SDR or USRP. The proposed blocks analyze the autocorrelation of samples through different methods to determine if they come from either noise or signals transmitted by communication devices. The reason for using the autocorrelation is that this feature is different for noise and for communication signals regardless of the noise power, which is an advantage in comparison with energy detection, another common method for spectrum sensing, which requires knowledge of the noise level to determine the presence or absence of communication signals. The first method consists in calculating the Euclidean distance between the autocorrelation and a reference line defined by the maximum values taken by the autocorrelation with high signal to noise ratio samples. To perform spectrum sensing with this method the proposed block takes the samples, calculates the autocorrelation and its distance to the reference line; if the distance is above certain threshold the block decides that the samples are signal, otherwise that the samples are noise. The second method consists in clustering the points defined by the variance and the mean of the autocorrelation samples. For implementing this method we identify one cluster containing samples with only noise, and other clusters containing samples from different communication signals. To put into practice this method, the block takes samples and estimates to which cluster they belong. The third method is based on the percentiles (25%, 50%, 75%) of the autocorrelation fast Fourier transform, better known as the power spectral density (PSD). For this method, we calculate the percentiles of the PSD of noise and communication signals at different frequencies and labeled them accordingly. We use the percentiles and their corresponding labels to train a KNN (K- nearest neighbor) classifier and create a classification model. To apply this method the proposed block takes the received samples, calculates the percentiles of their PSD and feeds them to the KNN model, which decides whether the samples come from either noise or communication signals. We implemented and tested experimentally the aforementioned methods with GNU radio companion, HackRF one, and RTL-SDR 2832u. In the talk we will present details about the proposed methods, their performance evaluation and the experiments conducted during the process.