With the rapid growth of satellite communications in various fields, particularly in military applications, and the high importance of communication security, the use of artificial intelligence-based methods to prevent the interception of communication signals and mitigate interference has become increasingly important. This paper proposes a multilayer perceptron (MLP) neural network-based method for the detection and identification of interference in the DVB-S2 satellite communication standard. The dataset used in this study includes modulated signals (QPSK and APSK with orders 8, 16, and 32) and various types of radio-frequency interference, including chirp interference (CI), continuous-wave interference (CWI), and multiple continuous-wave interference (MCWI). In the proposed method, a large number of statistical and spectral features are initially extracted from the signals. Then, the ten best features are selected using the Mahalanobis distance, Euclidean distance, chi-square, and variance-to-mean ratio (VdM) criteria to optimize the model performance. Numerical results demonstrate the improved accuracy of the VdM criterion in identifying interference, particularly under low signal-to-noise ratio (SNR) conditions. At SNR = 0 dB, the proposed method provides an accuracy improvement of up to 6% compared with previous methods, while at SNR = 9 dB, the accuracy of the proposed method reaches 99.74%.