Abstract: The traffic of the integrated space-terrestrial network has characteristics such as self-similarity and heavy-tailed distribution, which makes it difficult to efficiently distinguish abnormal traffic in the network. In view of this, we propose a hybrid detection method of abnormal traffic in the integrated space-terrestrial network based on entropy-optimized GA_LSTM. The method utilized information entropy to generalize the distribution of traffic characteristics to initially detect abnormal traffic, which could quickly narrow the scope of abnormal traffic detection. The genetic algorithm was used to optimize the LSTM and make a secondary judgment on the network traffic after entropy detection, which was used to improve the accuracy of abnormal traffic detection in the network. The simulation results show that the proposed traffic hybrid detection has better convergence speed and accuracy compared with traditional information entropy and classical LSTM detection algorithms.