Hybrid architecture for mitigating DDoS and other intrusions in SDN-IoT using MHDBN-W deep learning model

The Internet of Things (IoT) connects billions of devices. However, because of its heterogeneous system and broad connectivity, it is vulnerable to various intrusion challenges, resulting in data and financial loss. The IoT environment must be secured from such threats. This research proposes an SDN-enabled Deep-Learning-Driven System for IoT intrusion detection. Intrusion detection can detect unknown threats from network traffic and is a good network security measure. Most current network anomaly detection approaches use standard machine learning models like KNN and SVM. These approaches have some significant advantages, but they are not very accurate and rely on manual traffic design, which is outmoded in the age of big data. Our proposed Hybrid Deep Learning-based Intrusion Detection System (HDLIDS) addresses low accuracy and feature engineering issues. HDLIDS uses a novel Modified Hybrid Deep Belief Network with Weights (MHDBN-W) algorithm to detect existing and new cyberattacks. The MHDBN-W method consists of an MCL, a layer combining the MGBRBM and DNN-W algorithms, and an aggregator layer. The MHDBN-W technique has two phases: UL and SL of traffic features into normal and abnormal classes. The HDLIDS model is evaluated on the CICIDS2018 dataset compared to other conventional learning methods. It outperforms all other models in all performance criteria.