With the complexity and diversity of navigation spoofing signal attack strategies, how to effectively eliminate spoofing interference signals becomes very important. Based on (Empirical Mode Decomposition, EMD) and (Support Vector Machine, SVM), it has been proved that it can effectively process mixed signals with noise to obtain real satellite navigation signals. However, the EMD-based decomposition algorithm is prone to modal aliasing problems when decomposing signals, thereby reducing the suppression performance of deception signals. Therefore, this paper proposes a deception jamming signal suppression method based on complementary ensemble empirical mode decomposition and long-term and short-term memory network combination model. The method uses (Complete Ensemble Empirical Mode Decomposition, CEEMD) to decompose the received mixed navigation signal into IMF components with different characteristics, and establishes LSTM network models for the decomposed subsequences to predict. Finally, the real signal is accumulated according to the prediction results to eliminate the influence of deception signals. Theoretical analysis and simulation results show that compared with the spoofing interference suppression algorithm based on subspace orthogonal projection, this method can effectively eliminate induced spoofing interference, and its performance is better than the traditional EMD-SVM scheme. In addition, the algorithm eliminates spoofing signals at the signal level and has certain versatility.