Using Naive Bayes with AdaBoost to Enhance Network Anomaly Intrusion Detection

Classical intrusion detection system tends to identify attacks by using a set of rules known as signatures defined before the attack, this kind of detection is known as misuse intrusion detection. But reality is not always quantifiable, and this drives us to a new intrusion detection technique known as anomaly intrusion detection, due to the difficulties of defining normal pattern for random data frames, anomaly detection suffer from false positives, where normal traffic behavior is mistaken and classified as an attack and cause a great deal of manpower to manual sort the attacks. In this paper we construct a network based anomaly intrusion detection system using naive Bayes as weak learners enhanced with AdaBoost (Adaptive Boosing machine learning algorithm), experiment using KDD ’99 cup data proved that our IDS can achieve extremely low False Positive and has acceptable detection rate.