Comparitive Study of Intrusion Detection Using Hyper Heuristic Improved Particle Swarm Optimization and Hyper Heuristic Firefly Algorithm Based Convolutional Neural Network

SVM has high impact on big data classification problems but it requires expert knowledge for determining the configuration. It includes the selection of parameters and kernel functions which have the ability of increasing the classification accuracy. Therefore, configuration process of SVM is modelled here as multi-objective optimization problem with the parameters responsible for model complexity, false positive rate and false negative rate. The Hyper-Heuristic Improved Particle Swarm Optimization (HHIPSO) methodology is developed for optimizing the multi-objective problem of SVM. It is the fusion of hyper-heuristic and improved PSO techniques. The second model proposed is HHFA-CNN model used for detecting the intrusions which learns both the host and network level features. The intrusion detection performance is enhanced by selecting optimal configurations for CNN using HHFA. Finally appropriate configuration of SVM is detected with the assistance of HHIPSO. The benchmark dataset namely NSL-KDD is utilized to assess the effectiveness of HHIPSO model and HHFA-CNN model. In this paper we are going to compare the results between Hyper heuristic Improved Particle Swarm Optimization (HHIPSO) and Hyper heuristic Firefly Algorithm based Convolutional Neural Networks (HHFA-CNN) on the same data set.