A Deep Learning Algorithm to Cybersecurity: Enhancing Intrusion Detection with a Hybrid GRU and BiLSTM Model
Ameer A. Ghani, Suad A. Alasadi · Engineering Technology & Applied Science Research · 2025
Cyber security in networks and Internet of Things (IoT) environments is becoming complex with the evolution of sophisticated cyberattacks, and the existence of effective Intrusion Detection Systems (IDSs) is necessary. This work proposes a Network-based Intrusion Detection System (NIDS) for a hybrid Deep Learning (DL) model with Gated Recurrent Units (GRU) and Bidirectional Long Short-Term Memory (BiLSTM) to improve attack detection and classification. Pre-processing of datasets, feature selection with Pearson Correlation Coefficient (PCC), and training-testing with two benchmark datasets, CSE-CIC-IDS2018 and ToN_IoT, were performed. Surpassing standalone GRU and Bidirectional Long Short-Term Memory (BiLSTM) systems, the proposed hybrid model detected 99.86% of the attacks of the ToN_IoT dataset and 98.69% of the CSE-CIC-IDS2018 dataset while maintaining high accuracy, recall, and F1 score of over 99%. These results confirm that the proposed model can effectively counter traditional NIDS weaknesses through accuracy improvement in detections and with diversity and dynamics in networks' complex trends and IoT environments.