Abstract: Machine learning plays an important role in intrusion detection. As a key preprocessing step of machine learning, feature selection has attracted the attention of many researchers. Aiming at solving the problem that sparrow search algorithm has strong optimization ability but easy to fall into local optimization, this study improved the strategies of feature encoding and location update. A multi strategy mixed binary sparrow search algorithm was proposed, which combined a decision tree classifier to construct wrapped feature selection algorithm and selected representative features from high-dimensional feature space to improve the prediction ability of the model and reduce the time cost. The performance of this algorithm was evaluated based on NSL-KDD and UNSW-NB15 datasets. The experimental results show that compared with various feature selection algorithms, the data after feature selection using the algorithm proposed in this work has the best binary classification effect.