Enhancing Network Security on a Hybrid LSTM-Gradient Boosting Framework for Intrusion Detection

Abstract

The chapter presents a new hybrid intrusion detection framework that combines LSTM networks with gradient-boosting techniques. The approach presented in this study utilizes two prominent datasets, namely CICIDS2018 and CICIDS2017. The main objective is to improve the precision and reliability of intrusion detection in systems. This is achieved by capturing temporal dependencies in network traffic data and improving predictive performance by implementing boosting algorithms. The datasets are subjected to a thorough preprocessing process involving cleaning, normalization, and feature selection. This guarantees that the input for the model is of the highest quality. The LSTM component functions as the central element, extracting complex patterns and relationships from sequential data. The predictions generated by the LSTM model are refined by gradient boosting algorithms, which utilize their ensemble learning capabilities to improve overall performance.