DeepGuard: A Hybrid RF-CNN-GRU Framework for Adaptive Real-Time Zero-Day Network Intrusion Detection

Due to the dynamic and ever-increasing nature of cyber threats, the traditional network intrusion detection systems (NIDS) are not sufficient and are slow to respond to the adapting of new threats since they utilize rigid rules and signature-based detection techniques. This research study aims to design an intelligent adaptive intrusion detection system capable of precise and prompt detection of both established and latent threats to the network. Achieving optimal results from feature extraction, spatial pattern recognition, and temporal sequence learning requires a hybrid model that synergistically combines Random Forest, Convolutional Neural Network, and Gated Recurrent Unit (GRU). The model was built in Python using TensorFlow and Scikit-learn, training and testing it with the KDD Cup 1999 dataset. The experimental results indicated that the proposed model outperformed existing models, achieving an accuracy of 99.23%. This finding confirms the effectiveness of integrating multiple deep learning models. The research data illustrates how the models effectively resolve a basic cybersecurity challenge during active performance.