HYPER-Q: Guaranteeing the privacy of hybrid quantum learning through a theoretical mechanism

HYPER-Q is a noise-adding mechanism that mixes classical and quantum noise, used to protect the privacy of quantum machine learning models. Under the differential privacy framework, quantum noise combined with classical mechanisms can satisfy formal (ε, δ)-DP guarantees, thereby reducing classical perturbation without sacrificing the privacy budget, with the potential to improve utility. The authors provide privacy guarantees and theoretical utility bounds, and show on multiple real datasets that HYPER-Q's adversarial robustness is better than existing classical noise-adding mechanisms.