Network-Shield: Exploring the Efficacy of GRU Model in Intrusion Detection Using CIC-IDS 2018 Dataset

In today's interconnected digital landscape, protecting network systems from malicious intrusions is vital.Despite the increasing prevalence of cyber dangers, it is crucial to have strong and reliable intrusion detection systems.In this work we investigates the effectiveness of utilizing deep learning, namely the Gated Recurrent Unit, for network intrusion detection using the CIC-IDS2018 dataset.We have included all classes from the dataset in our study, addressing the often-overlooked minor classes typically excluded in similar research.The study concentrates on identifying minor attacks that, although not common, represent considerable security threats.In order to address the issue of imbalanced datasets, we utilize the Synthetic Minority Over-sampling technique, which improves the accuracy of the model.In addition, the study examines the application of the Random Forest algorithm for selecting features, hence enhancing the effectiveness of detection.Our proposed method demonstrates exceptional performance in network intrusion detection, achieving an impressive accuracy of 96.23% along with remarkable precision and recall scores.Through the utilization of a condensed feature set, the model optimizes computational efficiency without sacrificing precision, rendering it a pragmatic and potent solution for real-world cybersecurity tasks.Furthermore, it achieves this objective with fewer parameters and reduced storage requirements, affirming its practical utility in cybersecurity applications.