Should be merged with https://lobste.rs/s/rs9tke/hertzbleed_attack (or the other way around).
Okay so power consumption changes with transistor toggle rate. The toggle rates are all data dependent. Can this be worked around by adding blinding, like you already have to do to stop RSA leaking everything?
edit: Also I would suggest linking to the site with the paper https://www.hertzbleed.com/
Intel’s guidelines mention a range of mitigation options. I’m not sure about their efficiency – I just saw this and I’m still going through the paper (at a pace akin to that of a heavily-anaesthetised drunk turtle, because of burnout and lack of time ¯\_(ツ)_/¯) – but they seem sensible.