There exist some issues in current intrusion detection algorithms such as unbalanced detection performance on different types of attacks, and redundant or useless features that will lead to the complexity of detection model and degradation of detection accuracy. This paper presents an ensemble approach to intrusion detection based on improved multi-objective genetic algorithm. The algorithm generates the optimal feature subsets, which achieve the best trade-off between detection rate and false positive rate through an improved MOGA. And the most accurate and diverse base classifiers are selected to constitute the ensemble intrusion detection model by selective ensemble approach. The experimental results show that the algorithm can solve the feature selection problem of intrusion detection effectively. It can also achieve balanced detection performance on different types of attacks while maintaining high detection accuracy.