The paper aims at the insufficient of traditional intrusion detection based on genetic neural network not consider the misclassification cost, integrate the misclassification cost-sensitive features into the network intrusion detection model which based on genetic neural network, to overcome the defect of the traditional model's error classifying result in excessive costs. The experiment results show that after the genetic neural network increased the misclassification cost-sensitive features, it can control the cost caused by the network intrusion detection's false report、 omit report attacks preferably.