With the growing use of the internet worldwide, internet security becomes more and more important. There are many techniques available for intrusion detection. However, there remain various issues to be improved, such as detection rate, false positive rate, memory overhead, time overhead, and so on. In this paper, a new hybrid system for network intrusion detection system using principal component analysis and C4.5 is presented, which has a good detection rate and keeps false positive and false negative rate at an acceptable level for different types of network attacks. Especially, this system can effectively reduce the memory overhead and the time overhead of building the intrusion detection model. These claims are verified by experimental results on the KDD Cup 99 benchmark network intrusion detection dataset.