A Hybrid Light Gradient Boosting Approach with Deep Boltzmann Machine for Intrusion Detection System

Abstract

The need for network attack analysis and the number of cyber threats and attacks are increasing. Due to the scalability and adaptability of its internet-based computing resources, cloud computing is gaining popularity among businesses worldwide. Scientists are increasingly interested in cloud data security and face the difficulty of protecting hosts, enterprises, and data against more advanced digital threats. Over the past few decades, researchers have experimented with the Intrusion Detection (ID) paradigm, leading to various methodologies. However, critical to analyze the abnormalities in the intrusion detection framework. This research aims to classify the attacks or abnormalities in the network included in the NSL-KDD dataset using a Hybrid Light Gradient Boosting approach with Deep Boltzmann Machine. A Deep Boltzmann Machine Classifier and an ensemble model of the Light Gradient Boosting technique were used to create this model. Gradient boosting techniques improve the performance of DL classifiers by minimizing the number of errors discovered during network intrusion detection. This work evaluates and contrasts the proposed classifier with recognized classification strategies. The proposed model surpasses previous Recall, F-Measure, Precision, and Accuracy techniques.