Network intrusion detection is a hot research topic in network security, in order to improve the accuracy of network intrusion detection, a network intrusion detection model (IPSO-SVM) is proposed based on improved particle swarm optimization algorithm and support vector machine to solve the problem of classifier's parameters optimization. Firstly, network intrusion detection rate is taken as the objective function, and support vector machine parameters are used as the constraint conditions to establish mathematical model, and secondly improved particle swarm optimization algorithm is used to find the optimal parameters, finally, support vector machine is used as classifier to build intrusion detection model, and KDD 1999 data is used to validate the performance in Matlab 2012. The results show that IPSO-SVM has solved the optimization problem of the classifier's parameters and improved detection rate, reduced false alarm rate, false negative rate of the network intrusion.