Hybrid Deep Learning Approach for Efficient Intrusion Detection Using DCNN and BiLSTM

With the rise of digital connectivity, modern computer networks are increasingly vulnerable to dynamic and complex cyber threats. In response to these evolving risks, this study presents Deep Hybrid NIDS, an advanced Network Intrusion Detection System (NIDS) that leverages a synergistic deep learning framework integrating Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks. The proposed model is trained and validated using traffic data from two widely adopted benchmark datasets-CICIDS2018 and Edge-IIoT. This architecture enhances both feature extraction and temporal dependency modeling. Notably, the model incorporates an attention mechanism to prioritize salient features during classification, improving detection accuracy. Experimental evaluation demonstrates a high detection capability, with accuracy scores reaching 99.64 % on the Edge-IIoT dataset and near-perfect performance on CICIDS2018, outperforming several existing methods. The attention strategy and hybrid design constitute the core novelty, offering an effective solution for real-time intrusion detection in dynamic network environments.