Towards Quantum-Enhanced Intrusion Detection in Industrial Control Systems: A Proof-of-Concept Using Variational Quantum Classifiers

Intrusion detection in Industrial Control Systems (ICS) remains a significant challenge due to strict real-time constraints and evolving cyber threats. Although deep learning models like autoencoders have shown promise, they often require large datasets and are limited in their ability to generalize across scenarios. In this paper, we propose a novel intrusion detection method based on Variational Quantum Classifiers (VQC). By leveraging quantum circuits for feature encoding and classification, our method demonstrates strong anomaly detection capabilities, even with limited data. We evaluate the VQC model on a subset of the WUSTL-IIOT-2018 dataset and compare its performance to a classical Conditional Variational Autoencoder (CVAE) model. The VQC approach achieves $99.00 \%$ accuracy, perfect recall, and 0.9947 AUC-ROC, showcasing its potential as a lightweight and accurate anomaly detector for resourceconstrained ICS environments. This work provides one of the first explorations of quantum machine learning for ICS security and outlines key directions for future research on real quantum hardware and larger-scale deployments.