Improved Binary Chemical Reaction Optimization with Optimal Convolutional Long Short Term Memory Based Intrusion Detection System on Big Data Environment

Abstract

At recent times, the generation of massive amount of data and its continual raise have transformed the significance of data security and analytics systems for big data. To achieve security in big data environment, intrusion detection system (IDS) finds useful which observes and investigates the data for the detection of intrusions exist in the network. The existence of high volume, speed, and variety of big data has made it infeasible to employ the traditional IDS. Therefore, this paper presents a new hybrid metaheuristic optimization with deep learning based intrusion detection system (HMODL-IDS) in Apache Spark big data environment. In addition, the presented model involves preprocessing, feature selection, classification, and parameter optimization. Moreover, an improved binary chemical reaction optimization (IBCRO) algorithm as a feature selector to derive a useful set of features. The IBCRO algorithm is designed by incorporating the BCRO algorithm with Niche mechanism to avoid the local optima problem of BCRO algorithm. Moreover, the hybridization of convolutional neural network with long short term memory (HCNN-LSTM) model is employed to classification. At last, the grasshopper optimization algorithm (GOA) is applied to tune the hyperparameters of the HCNN-LSTM model on Apache Spark Big Data platform. An extensive set of experimental analysis take place on benchmark dataset and investigate the results interms of different measures.