Modified Firefly Algorithm based Optimum Feature Selection and Ensemble Tree based Model for Network Intrusion Detection using Data Mining Technique

Abstract

Intrusion Detection is the practice of recognizing items or events that do not follow an expected behavior or do not coordinate with other normal items in the dataset. Network traffic is increasing identifiable event to growing use of the web services and smart devices. The NSL-KDD is widely utilized dataset in the analysis of Intrusion Detection over computer networks. The dataset contains high dimensional data and also the imbalanced class. Due to this kind of dataset the imbalanced classification problem arrives. To overcome the deficit of data instances in one particular class, create extra data samples on that minority class. Detection of network anomalies from high dimensional dataset is critical and taking too much of time to process, so it is carry out using bio inspired feature selection technique. In the proposed system, the synthetic minority over-sampling Technique is used, which is one kind of effective method to rectify the class imbalance problem. Then the bio-inspired based features selecting process is carried out using Modified FireFly Algorithm (MFFA) and the resultant optimized dataset is taken for further process. After the features selection, the obtained dataset is fed into tree based J48 algorithm for build the Intrusion Detection System and detect the normal and anomalies in the network. Then, the ensemble bagged J48 classification is performed to improve the prediction accuracy.