Enhancing Network Security Intrusion Detection and Real-Time Response With Long Short-Term Memory Networks

In cybersecurity, network security systems (NSS) are technologies used to protect sensitive data against increasing cyber-attacks. This paper has carried out the process of integrating advanced Machine Learning (ML) techniques such as the Long Short-Term Memory (LSTM) networks along with the Convolutional Neural Networks (CNN) for the task of enhancing the Intrusion Detection Systems (IDS) in the NSS. Usually, the traditional models have faced many challenges, such as high false positive rates (FPR) and the need for real-time processing of extensive data streams, which make these systems insufficient to handle such scenarios. So, to handle these complications, ML has evolved to propose improvements in IDS implementation by adapting to new attacks and detecting complex patterns with greater accuracy. The study introduced a novel deep learning (DL), “GC-SLSTM,” that combined models such as Gated CNN with Stacked LSTM to address these challenges. This model includes CNN robust spatial pattern recognition and the LSTM that effectively handles the temporal data analysis. The proposed model was experimented with using the CICIDS2018 dataset, and the results demonstrate that the proposed Gated CNN + Stacked LSTM (GC-SLSTM) had achieved an accuracy of up to 99.59%, precision of 99.58% and a recall of 99.47%, culminating in an F1-score of 99.59%.