Classification of DDoS in Software-Defined IoT Networks using Hybrid Feature Selection Technique with DL

Abstract

A new network technology called Software Defined Networking (SDN) enables centralised network management. The SDN's potential features improve network security and make it easier to create threat detection systems through software programmes that make use of open APIs. However, the new technology generates dangers and security issues that are not present in the traditional networks that are now in use. One of the most frequent assaults that can stop the operation of the network and render the majority of network services inaccessible to customers is a distributed denial of service attack (DDoS). Because so many network variables are considered and machine learning-based anomaly detection methods must be used, it is still difficult to identify DDos assaults on SDN setups effectively. In order to analyse the most thorough significant elements of DDoS attacks in SDN networks, this research planned to apply feature selection methods Information Gain (IG), Random Forest (RF), and Embedded feature model. The accuracy of the attack detection system will be increased, and the incidence of false alarms will be decreased, by using the most pertinent features. To address the issue of DDoS attacks in SDNs, a Deep Learning (DL) technique based on Convolutional Neural Network (CNN) is also put out. The study analyses and evaluates InSDN, measures the effectiveness of the suggested DL model on the SDN controller, and tests the performance of the network in terms of network throughput and end-to-end latency. The outcomes demonstrate that the DL technique effectively detects DDoS attacks in SDN settings without materially affecting controller performance.