Various machine learning techniques have been used for network intrusion detection. The supervised machine learning methods can achieve high accuracy in classifying normal and abnormal network data. However, a large amount of labeled data is needed to acquire high accuracy. Labeling large amount of data could be very costly. Semi-supervised machine learning techniques overcome this problem since they only use a small amount of labeled data and large amount of unlabeled data.

In this research, semi-supervised Support Vector machine (SVM), Random Forest and Deep Belief Network (DBN) were used in classifying network data for intrusion detection. They were used to classify the Third International Knowledge Discovery and Data Mining Tools Competition dataset (KDD 1999). The results of semi-supervised Random Forest for classifying normal and abnormal network data were compared with the results of using supervised Random Forest. The results were also compared with semi-supervised ladder network in classifying KDD 1999. Self-learning based semi-supervised Support Vector Machine (SVM) and Deep Belief Network (DBN) were also used to classify the specific attack types in KDD 1999.