A hybrid deep learning approach with temporal awareness for intelligent intrusion detection in 6G-enabled IIoT networks

Abstract The integration of the sixth-generation (6G) communication technology and the Industrial Internet of Things (IIoT) has realized the intelligence and automation of industrial applications. However, due to the complexity, dynamics, and heterogeneity of data, traditional threat detection methods make it difficult to deal with cyber threats in the 6G-IIoT environment. In view of these limitations, this study proposes a hybrid Deep Learning (DL) model combining a Deep Neural Network (DNN), a Bidirectional Gated Recurrent Unit (BiGRU), and an attention mechanism for threat detection in a 6G-IIoT environment. DNN extracts global features, BiGRU captures bidirectional temporal dependencies, and the attention mechanism highlights key anomalies. Experimental results on the Edge-IIoTset dataset show that the accuracy rate of the model is $$96.88\%$$ . It outperforms baseline models (such as ANN, CNN, DNN-LSTM). The model achieves high accuracy and low False Positive Rate (FPR), and meets the dynamic security requirements of the 6G-IIoT environment. This research provides a promising solution for real-time threat detection in next-generation industrial networks.