NIDF: An Ensemble-inspired Feature Learning Framework for Network Intrusion Detection

In today's world growth of network-based applications increasing rapidly, to protect sensitive information from various threads and attackers an effective intrusion detection system needs to be developed. A network administrator completely depends on an efficient intrusion detection system that detects intrusion from the network. So to make our information safe and secure we need to detect the intrusion such that the hackers are prevented from the damage our information. The machine learning-based intrusion detection is one of the useful approaches to detect the intrusion present in the network. In this article, we have developed an ensemble-inspired feature learning environment for detecting intrusion in the network using several machine learning based classifiers. At first, we have identified the furthermost suitable features from the NSL KDD dataset using multiple feature selection methods like gain ratio, relief, and information gain strategies. After that, the top most relevant feature has selected by using our ensemble method from the combined pre-identified features set. The robustness of our model has been measured by applying k-fold cross-validation. The classification accuracy we have measured by applying several machine learning classifiers (J48, Decision Tree, and Random Forest). We have seen that the random forest classifier gives the best classification accuracy of 99.58%. Finally, we have analyzed that our ensemble method gives the best result compared with other feature selection methods by using same of classifiers.