A novel and an effective intrusion detection system using machine learning techniques

Network environments become more and more diverse with the presence of many different network protocols, services, applications and so on. With this diversity, many different types of attacks appear and target at a computer or a network every day. A single type of intrusion detection systems (IDSs), which has its own advantages and disadvantages, seems to be insufficient to detect all the attacks. Since us don’t know which types of attacks are coming next, the primary difficulty lays on selecting of the best IDS at a certain time. In our scenario, we assume that each IDS has its own favorite types of attacks to detect. In this paper is investigated for intrusion detection system (IDS) and its performance has been evaluated on the normal and abnormal intrusion datasets (KDDCUP99). New technique of k-NN algorithm using NA (Network Anomaly) rules for intrusion detection system is experimented. The research work compares accuracy, detection rate, false alarm rate and accuracy of other attacks under different proportion of normal information. Comparison between Naive Bayes classifier, SVM and NA-kNN for same training data set and testing data set has carried out. Experimental results show that for Probe, U2R, and R2L, NA-kNN gives better result. Overall correct count to detect correct attacks is larger in NA-kNN than other classifier algorithms.