ML-Based Synthetic OPC UA Traffic Generation for Covert Channel Transmission

This paper presents a novel approach for generating synthetic OPC UA traffic using a hybrid model that combines Variational Autoencoders (VAE) and Long Short-Term Memory (LSTM) networks to address the cybersecurity challenges in industrial automation systems. The Open Platform Communications Unified Architecture (OPC UA) is widely adopted for its versatility and secure data exchange capabilities across various sectors. However, the sensitive nature of OPC UA data traffic presents significant challenges in cybersecurity, particularly concerning covert channel attacks. These attacks exploit legitimate communication protocols to stealthily transmit information, leading to potential data breaches and unauthorized access. By creating realistic, privacy-preserving synthetic data, the model facilitates the analysis and detection of covert channel attacks—stealthy methods used to transmit unauthorized information through legitimate communication protocols. This approach not only provides a valuable dataset for researchers but also supports the development and evaluation of robust mitigation strategies without the need for sensitive real-world data. Demonstrating the model's effectiveness through the successful embedding of covert messages, this work paves the way for enhanced security measures in OPC UA networks and contributes significantly to the field of industrial cybersecurity.