Current setup: I click on bad link, malware installed on my computer that infects my disk firmware. I remove the malware from my computer. My disk remains compromised.
Suggested setup: I click on bad link, malware installed on my computer that infects my disk firmware. I remove the malware from my computer. My disk is no longer compromised.
Also, I plug a USB stick into my computer. It gets compromised. I unplug it. It gets uncompromised. No more USB spread malware.
Your suggested setup would allow a worm to repeatedly reinfect everything. Some forms of malware, such as one used in an attack against Kaspersky Lab, have been designed to do so instead of persisting directly onto the targeted devices.
The exponential growth of the infection will be either blunted or stopped if a significant percentage of the infected machines keep getting uninfected. It's "herd immunity".
Not that I know of. I suggest it every time the subject comes up of hardening internet connected devices. Nobody has ever replied with an instance of it.
Back in the 80's, most devices used ROMs, PROMs or EPROMS. ROMs were burned at the factory. PROMs could be user programmed once. EPROMs were erasable and reprogrammable, if you put the chip under a UV lamp. As far I can tell, this is a forgotten technology.
Later on came EEPROMS, electrically erasable programmable read only memory. Then someone had the bright (!) idea of connecting the write-enable line to the internet so anybody in the world could update anybody else's firmware, and welcome to the hell we have today.
Option B) Code in PROM with physical interlock (e.g. push button to enable write)
Option C) Code in ROM
Downsides of A are obvious, we see them now.
Downside of B is that automatic updates are impossible, so the majority of devices that are network connected will remain vulnerable. At the peak of code red, it took less than 5 minutes for a fresh windows install to be compromised.
You also will have people who will forget to disable the network before pushing the button, or who may be tricked into pushing the button which will allow malicious code to persist anyways.
C) Absent a hardware recall, people will just use devices that are infected within 5 minutes of bootup.
This implies a lot of infected machines. As I mentioned earlier, machines getting uninfected at every reboot will reduce the number of infected machines at any point in time. It may reduce it far enough to provide "herd immunity", which is why we only need 90% of the population immunized against measles to prevent it from propagating.
Furthermore, if people are aware that reboots will de-infect malware, machines can be set up to regularly reboot. For example, I could set up my router to reboot once an hour. This would be only a minor inconvenience, as it reboots pretty quickly.
Regular reboot can be done by adding one of those simple hardware store lamp timers, or the device itself can contain a hardware circuit (outside of software control) to regularly reboot it.
By the way, do your really want automatic updates to your disk drive firmware? Your USB stick firmware? I don't. They don't happen anyway, yet I've seen articles about how those get infected with malware.
Early PC had all their OS in Rom. Apples (Mac, etc) for instance. It was abandoned because it was difficult to patch the OS. The patch were needed to fix unpleasants "features" of the program. The back side is that it also allows to install backdoors or snooping extensions.
A better approach would be to create simpler systems that are easier to verify and monitor, and harder to corrupt. Another approach is to use micro kernel architecture and a container oriented system.
Finally, the backdoor strategy of the NSA is now proven to be harmful. With all their tools, they should know who is behind it.
