Research on Intrusion Detection Method Based on SVM Co-training

Currently, network intrusion detection is in face of the conflict between the difficult to label data and the high accuracy request to detect intrusion. In this paper, we propose a SVM co-training based method to detect network intrusion. It exploits the large amount of unlabeled data, and increase the detection accuracy and stability by co-training two classifiers. The simulation results show that our method is 11.9% more accurate than the traditional SVM method, and it depends less on the training dataset and test dataset.