Securing Smart Systems: A Parallel Orchestrated LSTM-CNN Model for Cyber Threat Detection in Industry 4.0

Protecting Industry 4.0 networks against sophisticated cyberattacks is the only option by which industrial automation and intelligent manufacturing systems can be trustworthy, stable, and resilient. The increased count of IoT, IIoT, and cyber-physical elements installed within these systems makes them vulnerable to new and sophisticated categories of threats. Existing detection methods are typically non-real-time, non-adaptive, and imprecise mainly because current systems do not completely digitize temporal and spatial trends within industry data streams. In this scenario, this paper proposes an innovative deep learning-based technique known as Parallel Orchestrated LSTM-CNN Network (PLO-LCNet) to efficiently detect cyberthreats in Industry 4.0 networks. The approach includes Parallel Sequential Anomaly Detection using Long Short-Term Memory (LSTM) and Parallel Spatial Feature Learning using Convolutional Neural Networks (CNN) in a parallel design, which is performed to improve the efficiency of learning. The architecture allows pattern detection as well as performance improvement simultaneously. In contrast with the conventional benchmarking datasets like Edge-IIoTset and CSE-CIC-IDS2018, the detection accuracy of PLO-LCNet is 0.99 and precision and recall accuracy is 0.98 and 0.99, respectively. Experimental results indicate that the model is better than other existing methods in aspects of accuracy, response time, and fewer errors and thus ideal for Industry 4.0 system security.