This is the fun part of VMI: the guest can lie, but the hypervisor does not have to believe it.

The painful part is translating that truth into useful detections without melting performance or creating a noise cannon.