[lifeisstillgood]

I wonder if it is time for a reboot. If the castles we have built so far turn out to be made of gauze instead of stone, maybe we need to rethink it all, in the same way we need to rethink energy policy

Every Intel motherboard since 2008 has had a "spy" on board, almost every home router is working for someone's botnet and will never be patched, medical devices and factory automation systems ship with default passwords because no one assumed they would ever connect to the Internet and don't get me started on browsers and JavaScript.

It was a multi-decade long fight to get the seat belt adopted, so I suspect that we aren't going to fix this the old way - surely at some point we stop?

[tmzt]

RiscV, TCP+crypto offload, hardware switchports with luajit or nf rules. Reactive UI with hardware rendering and compositing.

Hardware keystore with physical switch to generate and enroll keys, user/owner controlled secrets, one-time programmable as an option, hardwired SAK and OS personality switching key.

Real-time security isolation kernel, hardware-enforced containerization with MMU-protected GPU passthrough.

[lifeisstillgood]

It will take a while to google-walk through all that, but thank you. Do you feel this is a comprehensive recipie to move to a (enterprise wide) computing platform where the attacker has the paying field tipped against them (it seems the other way round today)

[Jach]

It doesn't sound comprehensive enough to me, though better than what's around. My own comprehensive recipe is simply "put nickpsecurity in charge". :)

[tmzt]

I was thinking the same thing. What I was describing is about using the disadvantages of a platform like RiscV yo our advantage. Rather than running network stacks, compositing and other things on the main processor which will likely trail intel processors in performance for a time, we design the hardware to do what hardware does best.