An ensemble of feature subset selection with deep belief network based secure intrusion detection in big data environment

Abstract

In recent times, massive quantity of data and its exponential rise have modified the significance of data security and analysis systems for Big Data. An intrusion detection system (IDS) is a process which helps to monitor and analyze the data for detecting the intrusions in the network. The huge quantity, variation, and high speed of data produced in the system pose a difficulty to the conventional techniques for the detection of attacks. Big Data methods are utilized in IDS for dealing with Big Data for precise and effective data analytic processes. This paper introduces a novel feature subset selection (FSS) with deep belief network (DBN) based intrusion detection in big data environment, called FSS-DBN model. The presented model involves data preprocessing stage to improve the data quality to a certain extent. In addition, the FSS can be considered as an optimization problem and is effectively solved by the use of teaching and learning based optimization (TLBO) algorithm. Moreover, DBN model is applied for identifying the class labels (i.e. intrusions) in the network. For validating the proficient results analysis of the FSS-DBN model, an extensive set of simulations were performed and the superior performance is also highlighted with the maximum sensitivity of 0.9898, specificity of 0.9865, accuracy of 0.9854, F-score of 0.9872, and kappa of 0.9867.