Hybrid Metaheuristics with Deep Learning Enabled Cyberattack Prevention in Software Defined Networks

Abstract

Software-Defined Networks (SDN) refers to a revolutionary pattern that separates the control plane from the data plane, converting the idea of a software-driven network. Cyber attackers had a target towards the SDN controllers to subdue the control planes that can be regarded as the SDN brain. It offers a plethora of functionalities like regulating flow control to routers or switches in the data plane below through southbound Application Programming Interfaces (APIs) and application logic and business in the application plane above through northbound APIs for implementing sophisticated networks. But the control plane is a tempting prospect for security attacks from adversaries due to its centralization features. The main concern is information safety in the network. To prevent the loss of extremely useful information, the Intrusion Detection System (IDS) has been formulated for recognizing the outbreak of a stream of attacks and notifying system administrators granting network security. With this motivation, this article develops a Hybrid Metaheuristics with Deep Learning Enabled Cyberattack Prevention (HMDL-CAP) model in SDN. The presented HMDL-CAP model initially carries out data preprocessing to scale the input data. Then,spiral dynamics optimization-based feature selection (SDOFS) algorithm is utilized for optimum selection of feature subsets. Next, hybrid convolutional neural network with recurrent neural network (HCRNN) model is applied to detect intrusions. As hyperparameter tuning is important, pelican optimization algorithm (POA) is used to tune the HRCNN parameters. To assess the experimental outcomes of the proposed model, a series of experiments were performed using benchmark dataset. The comparison study shows the promising performance of the HMDL-CAP model over recent models.

Keywords

• cyberattack prevention
• deep learning
• feature selection
• metaheuristics
• security
• software defined networks