Intrusion detection in big data using hybrid feature fusion and optimization enabled deep learning based on spark architecture

Abstract

With the rapid expansion of Internet services and increasing intrusion issues, conventional intrusion detection techniques cannot work well with several difficult intrusions. Though several intrusion detection methods have been introduced in the past years, developing an Intrusion Detection Systems (IDS) with prevailing intrusion detection strategy is still very much desirable. Hence, a robust and effective intrusion detection approach, named RV coefficient+Exponential Sea Lion Optimization-enabled Deep Residual Network (ExpSLO-enabled DRN) using spark is devised for the intrusion detection. Here, the unique features are selected using proposed RV coefficient-based hybrid feature fusion, which is designed by the incorporation of wrapper, class-wise information gain (CIG), and Canberra distance in slave node. With the distinctive features selected, the process of data augmentation is done using oversampling for making the data more appropriate to perform the further process in the slave node. Moreover, DRN classifier is utilized for detecting the intrusions in the master node where the DRN training is done using devised ExpSLO algorithm, which is the hybridization of Exponentially Weighted Moving average (EWMA) and Sea Lion Optimization (SLnO). Furthermore, the devised method obtained better performance by considering the evaluation metrics, such as precision, recall, and F-measure with the higher values of 0.8800, 0.8845, and 0.8822 based on without attacks using dataset-2.