A Hyper Meta-Heuristic Cascaded Support Vector Machines for Big Data Cyber-Security

Abstract

At an incredible speed, cyber security evolves in the ever-changing setting of attacks. Organisation processing of information inward and outward is huge in quantity and determining a threat amidst of information is challengeable. Late discovery of such instance is standstill challenge of the meticulous process. Thence, detection of intrusion and its prevention are rising challenge in Big data factors. the information inundation generally incorporate the Big data terms to dataset. The majorly focused issues are industrial oriented in big data challenge. Existing systems for big data cyber security problems are based on Online Support Vector Machines (OSVMs) framework. Bi-objective optimisation problem with primary objectives is designed as OSVMs configuration process for improving accuracy and less complexity of model. Here, a bi-objective optimization is implemented based on an Artificial Bee Colony (ABC). However, Online Support Vector Machines (OSVMs) has issue with computational complexity, and prematurity and local optimum is major problems in ABC algorithm. By overcoming this issue, developed research system designs an Ensemble Support Vector Machine (ESVM) framework for big data cyber security. Initially, the feature selection is done by using improved K-means clustering. Based on the selected features the intrusion detection and malware detection are performed using ESVM approach. In this proposed research work, a bi-objective optimization problem is designed as the ESVM configuration process for improving accuracy and less complexity of model and achieve its objectives. Cuckoo Search (CS) optimization algorithm is implemented for the bi-objective optimization. accuracy, precision, recall and f-measure are the parametric meters compared in proposed research attaining higher performance against existing approaches.