| An Optimized Multi-Class Prediction Model for Network Intrusion Detection Based on GRU |
| Paper ID : 1071-ICEEM2025 |
| Authors |
|
Eman Zakaria Mohamed * Benha faculty of Engineering |
| Abstract |
| Cybersecurity has become the most important priority in the information technology (IT) sector of any organization. Recent intrusion incidents have emphasized the crucial role of network intrusion detection systems in thwarting increasingly complex network attacks, particularly with rapid internet connectivity and traffic volume. Building efficient and broadly applicable security solutions still depends on determining the best learning paradigms for intrusion detection. This field in cybersecurity is considered very promising due to its importance. The goal of this work is to develop an efficient Multi-classification Network Intrusion Detection System (NIDS) using an optimized Gated Recurrent Unit (GRU). Our proposed system is validated using the CICIDS2017 dataset. The SMOTENN approach has been used to balance the dataset to improve the results. The performance assessment depicts that the proposed optimized GRU model based on a reduced feature set detects multi-class intrusions with a high overall accuracy of 99.42%, 99.40% precision, and 99.40% recall. |
| Keywords |
| Deep Learning, Gated Recurrent Network (GRU), Network Intrusion Detection System (NIDS), Multi-Classification. |
| Status: Accepted |