As cloud-hosted quantum machine learning models become standard, they increasingly face the risk of model stealing attacks. This study provides insights and potential defenses tailored to the realm of quantum computing.

Key Insights:

• Model stealing attacks on QNNs can yield highly-accurate clone models.
• Proposed defenses introduce perturbations using hardware variation.
• QML models trained on noisy devices may inherently resist common attack strategies.

An imperative emerges to develop robust safeguarding mechanisms for the nascent domain of QML models amidst model stealing threats. Delve deeper into the study.