Does quantum kernel alignment transfer across network intrusion datasets? Cross-dataset consistency, direct transfer and high performance computing cost of entanglement-based feature maps
This study measures two distinct notions of transferability for quantum support vector machines (QSVMs) across InSDN, CICIDS2017 and CIC-DDoS2019: cross-dataset consistency of kernel alignment (KA), and direct transfer of an InSDN-trained kernel classifier to the other two datasets without any target-side fitting.