Hybrid Intrusion Detection Technique for Malicious Network Attacks with Machine Learning

The number of malicious network attacks is increasing due to the corresponding increase in the number of Internet users. This trend has enabled attackers to exploit the vulnerabilities of network-connected devices to execute successful attacks. In this paper, a hybrid intrusion detection technique using machine learning algorithms is proposed. First, feature selection is performed to reduce the complexity of the model based on the size of the input. The selected features are used as input to the model for classification. To classify the observations as either anomalous or normal connections, pattern matching is carried out to match each instance of a network connection to a class label to optimise the performance of the classification algorithm. Furthermore, our model was evaluated using the KDD'Cup99 and NSL-KDD datasets, and achieved average accuracy and detection rate of above 99% with significantly low false positive and error rates as against comparative approaches.